Deep Learning Applications in the Resilience Critical Infrastructure Systems—A Systematic Review

Deep Learning Applications in the Resilience Critical Infrastructure Systems—A Systematic Review | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Deep Learning Applications in the Resilience Critical Infrastructure Systems—A Systematic Review H L Gururaj This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5923379/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Technological advancements like AI, blockchain, and IoT are merging to bring about a new level of digital change. Critical infrastructure systems (CISs) are vital to modern society, as they support crucial social functions, economic organization, and national defense. Recently, the resilience of CISs has garnered attention in academic and policy fields, particularly in light of increased natural and technological disasters. However, assessing CIS resilience remains challenging, particularly in its practical application to operational risk management. Integrating advanced technologies with critical infrastructure (CI) can significantly enhance the quality of life and boost national economic productivity. Nevertheless, the lack of robust cybersecurity in CI has given rise to advanced threats and vulnerabilities, undermining these potential benefits. The paper explores cyber vulnerabilities and dangers in various critical structures, including the financial, agricultural, energy, and health systems. Moreover, we examine the positive aspects of artificial intelligence and provide a rich taxonomy of solutions that show how well AI-based approaches deal with different types of cyberattacks on critical infrastructure. Deep learning AI Critical Infrastructure Smart Grid Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Introduction The global wave of digital transformation has opened new avenues for cyber attackers, making many individuals vulnerable to phishing and other malicious activities (Jadav et al., 2023 ). From small shopkeepers using UPI applications to supermarkets relying on mobile applications, digitalization has become pervasive. Unfortunately, this widespread adoption has increased cyberattacks targeting these digital platforms. CI is basically the backbone of economic growth and development in a nation's journey towards prosperity. The arrival of IoT has brought a sea change in traditional CI, making them more connected and intelligent systems. Such intelligent transformation easily coordinates and communicates between large chunks of CI, such as power grids, transportation networks, and healthcare systems, with a small amount of data exchange. However, IoT networks are often based on obsolete legacy protocols that degrade the operational performance of CI (Adil et al.,2022; Alyahya et al., 2022 ). This makes the device quite vulnerable because an adversary can easily mount a cyberattack that can range from man-in-the-middle attacks to DDoS and malware attacks. The integration of computer technology across various sectors has led to an increased risk of cyberattacks. These attacks have far-reaching impacts, affecting fields such as education, military, healthcare, communications, finance and banking, manufacturing industries, transportation, and software and hardware vendors, as illustrated in Fig. 1 . Technological advancements have driven the adoption of IoT across multiple domains, enabling easier monitoring, customization, and data updating. However, identifying and eradicating cyberattacks remains a formidable challenge. Researchers have proposed cryptographic solutions, such as authentication and authorization, to ensure that only authorized individuals can access critical facilities and prevent unauthorized access attempts. These solutions often involve verifying identities through passwords or biometrics (Annadurai et al., 2022 ). Artificial intelligence-based systems can effectively reduce the risks associated with security threats in ‘Critical Infrastructure contexts despite their huge limitations (Garabato et al., 2022 ; Lee et al., 2020). Several appraisal surveys or review materials illustrating the application of AI on security matters have been made over time. However, most of these surveys are irrelevant to the CI applications (Roopashree et al., 2022 ). Therefore, it is essential to research and analyze the scope of activities involving managing security and privacy issues in CI employing AI. With this, we propose – a systematic review that explores the state of affairs in AI in the context of cybersecurity events and issues and the prospects and barriers for AI in CI. Figure 2 displays the estimated financial losses in dollars from various types of cyberattacks. Malware is the most damaging attack, causing approximately $ 23.6 million in losses, followed by web-based attacks at $ 20.1 million. Denial of services, malicious insiders, and phishing & social engineering also contribute significant losses, ranging from $ 1.3 million to $ 1.6 million. Lesser but notable losses are attributed to malicious code, stolen devices, ransomware, and botnets, with botnets accounting for the most minor financial loss at $ 350,012. The chart highlights the need for robust cybersecurity measures to mitigate the economic impact of these prevalent threats. Figure 3 provides a statistical overview of cybercrime incidents across various sectors. These statistics highlight the growing threat landscape and emphasize the need for robust cybersecurity measures, particularly in critical infrastructures. While numerous studies exist on cybersecurity and AI, few focus specifically on AI-based cybersecurity for CI, as shown in Fig. 4 . Existing surveys and review articles primarily address security concerns but must include a comprehensive taxonomy highlighting AI approaches' intuitive nature in confronting these concerns. As Fig. 5 shows, most of the primary studies were published in ScienceDirect. Additionally, they often lack a proof-of-concept demonstrating why AI is a pivotal technology in cyberspace. Based on these observations, our proposed survey makes the following significant contributions: Present a thorough review of AI-based Cybersecurity in CI applications. This review covers critical sectors such as the financial sector (banking, insurance, and stock market), the energy industry (nuclear power plants & sustainable energy203), agriculture, public health, and relevant security features. We offer a more complex taxonomy as it draws solutions for such critical areas using various AI techniques. Finally, discuss emerging scientific areas that inhibit the progress of CI and do not have a resistive preventive approach. The paper is organized as follows: Section II explores various critical infrastructures, Section III examines the taxonomy of AI, Section IV discusses challenges and future opportunities, and Section V presents the conclusion. 2.1 Critical Infrastructure Integrating sensors and actuators across sectors and technologies is called Critical Infrastructure (CI). CI serves as the backbone for human existence in everyday life scenarios. The US government oversees a variety of CI within society, as illustrated in Fig. 6 . 2.2 Chemical Sector Cyberattacks against storage facilities can cause environmental disasters through sabotage or unlawful access to hazardous goods, as well as the theft of intellectual property and compromise of private data. The Chemical Sector is particularly susceptible to cyber threats due to the sensitivity of the items stored there and the potential for significant consequences on the environment and public safety. Storage facilities are vulnerable to these risks because they often handle dangerous materials, confidential data, and cutting-edge proprietary techniques. 2.3 Commercial Facilities Sector Cybersecurity in the corporate facilities sector has become more concerning due to the heavy reliance on digital technology in office buildings, retail stores, and other commercial spaces. Because these institutions use networked systems and intelligent technology to improve customer service and optimize operations, they are especially vulnerable to cyberattacks that target critical infrastructure. 2.4 Communications Sector It is evident that the telecommunication industry is vital in sustaining the world's communication infrastructure, which is essential for daily living, commerce, and defense. As these networks worldwide expand and become increasingly available, this industry continues to experience an expanding number of cybersecurity threats. This is particularly the case since ISPs regularly pass volumes of information, thus making the sector a lucrative target for hackers. Key concerns with the telecommunications industry include sensitive data management, increasing dependence upon satellite technologies, and operational difficulty in securing a global network backbone. 2.5 Critical Manufacturing Sector It forms a vital part of the world's economy, producing commodities that are necessities in all industries, from food to medicine, energy, and other types of material. On the other hand, as critical digital networks, automation, and IIoT devices grow in prevalence through manufacturing processes, this has slowly become the logical point for cybercriminals to penetrate the lucrative industry. Supply chain disruption, ransomware attacks, and data breaches are among the cybersecurity risks that the sector is most vulnerable to. 2.6 Dams Sector The dam sector is essential to the provision of services like agriculture, water production, flood control, and hydroelectric power generation. Cyberattacks on this sector could have serious consequences and jeopardize both the stability of the economy and public safety. Cybercriminals see the sector as a lucrative target because of its vast irrigation systems, power generation facilities, canals, and reservoirs. 2.7 Défense Industrial Base Sector In addition to research, design, development, manufacture, delivery, and maintenance services, the DIB Sector offers military technologies, equipment, and services to the armed forces. An industry that is crucial to national security, it includes all companies that provide services primarily to the armed forces as well as those that design and manufacture defense systems that include weapons, airplanes, and cybersecurity tools. The DIB sector is a highly high-value target for cyber espionage and sabotage due to its strategic importance. Strong cybersecurity is the means by which personal data is shielded and military preparedness is upheld. 2.8 Emergency Services Sector The emergency services sector is essential to maintaining public safety since it provides services including rescue and disaster response, emergency medical services (EMS), police, and fire. Due to its increased reliance on digital technology for communications, coordination, and data management, this industry is especially susceptible to cyberattacks. These risks pose significant threats to public safety because they may make it more difficult for emergency services to respond appropriately in circumstances where lives are at risk. 2.9 Energy Sector The functioning of modern society depends on the energy industry, which comprises essential infrastructure such as electric power networks, oil and gas pipelines, and nuclear power plants. Because the industry is vital to providing electricity to residences, commercial buildings, and transportation networks, it is especially susceptible to cyberattacks. These attacks can inflict significant interruptions in the energy supply, financial losses, and a cascade impact on daily life and economic stability. 2.10 Financial Services Sector The financial services industry, encompassing a wide range of enterprises like banks, credit card firms, insurance companies, and investment corporations, is the cornerstone of the global economy. This industry is particularly vulnerable to hackers because it handles sensitive data and financial transactions. Robust cybersecurity measures are crucial in averting financial crimes such as fraud and data breaches, which can potentially undermine public trust in the financial system and cause substantial economic damage. 2.11 Food and Agriculture Sector The food and agricultural sector includes various activities, including farming, food processing, and supermarket distribution networks. This enterprise, which manages the entire supply chain from production to consumption, is essential to preserving food security and public health. Cyberattacks targeting this sector could disrupt the food supply chain, posing a significant risk to public health and causing shortages and contamination. 2.12 Government Facilities Sector The government facilities sector includes many essential buildings and infrastructures, such as public administration buildings, military installations, intelligence agencies, and public service offices. The government, public administration, and national security depend on these infrastructures. Because of their significance, they are regularly the target of cyberattacks intended to steal sensitive information or obstruct government operations. 2.13 Healthcare and Public Health Sector The public and healthcare sectors are critical to maintaining and improving public health and safety. This sector includes hospitals, clinics, public health organizations, and research centers. In addition to managing massive numbers of sensitive patient data, these firms provide essential medical services. Cyberattacks against healthcare organizations have a high risk of endangering patient security and preventing essential medical services from operating as usual. 2.14 Data centers Modern society is built on data centers and telecommunications networks, providing a wide range of vital services for people and businesses. They oversee and maintain significant digital assets, such as communication networks and copious volumes of private data. Data centers are the main targets of cyberattacks because of their essential function and the importance of the information they store. These facilities must be protected to preserve national security, economic stability, and public safety. 2.15 Nuclear Reactors, Materials, and Waste Sector The Nuclear Reactors, Materials, and Waste Sector includes the generation of nuclear energy, the management of radioactive materials, and the removal of nuclear waste. Managing the radioactive materials and waste that result from using nuclear power to generate a sizable portion of the world's electricity is the responsibility of this industry. Because security breaches could have catastrophic repercussions, this industry needs the most significant cybersecurity protection to counter threats such as sabotage and unlawful access. 2.16 Transportation Systems Sector The nation's ability to move people and products depends on the transportation systems industry, which includes public transportation networks, highways, bridges, and airports. Given the significant effects cyberattacks on this company could have on trade, public safety, and national security, this industry's security and uninterrupted operations must be preserved. 2.17 Water and Wastewater Sector Sanitation, environmental protection, and public health depend on the water and wastewater sectors. Stormwater management, wastewater treatment, and water supply management are all included in this industry. Because the sector is vital to preserving the environment's health and the clean water supply, it is a popular cyberattack target. Strong cybersecurity protections are required because assaults can have serious negative effects, such as water contamination, supply interruptions, and environmental damage. Table 1 presents an overview of existing research on critical infrastructure across various sectors. It summarizes the methodologies employed in each sector, highlighting their advantages and limitations within specific domains. Table 1 Existing works on various sectors Ref Sector Methodology Pros Limitation (Met et al., 2022 ) Banking Using Turkish bank data, time series models, and Auto ML and clustering methods Scalable, improved decision-making, and efficiency Requires high data quality and reliability issues may arise (Nabipour et al., 2020 ) Stock market Comparison of models applied to 10 years of historical data with continuous and binary input Credibility of results Focuses only on technical aspects (Kochunas et al., 2021) Nuclear sector Discussion on Decision Trees (DTs) and their benefits in Nuclear Power Plants (NPP) operations Improved performance, optimized operations, testing, and validation Data quality, availability, and complexity issues (Blessy et al., 2021) Crop production Covers components, techniques, and parameters in crop production Optimizes water usage and reduces energy consumption Integration of sensors can be challenging and expensive (Hemming et al., 2019 ) Greenhouse production Experiment using various AI algorithms evaluated across five teams Efficient resource consumption Requires specialized knowledge and can face reliability issues (Xu et al., 2021 ) Smart healthcare Neural network-based approach aimed at minimizing error Enhances security Transparency and bias issues (Wazid et al., 2022 ) Smart healthcare Healthcare 5.0 incorporating AI, IoT, and Big Data Reduces operational costs Implementation can be expensive (Nazar et al., 2021 ) Smart healthcare Discussion of HCI, ML characteristics, challenges, and future scope Transparent methods and unbiased reviews Relevant studies not considered (Ale et al., 2019 ) Crop disease Proposed a DNN using IoT with reduced input image size and transfer learning to detect plant diseases Accurate with low computational cost Transparency of classification issues (Dhieb et al., 2020 ) Insurance XGBoost algorithms used to detect fraud claims and risks Provides accuracy and efficiency Real-world applicability and scalability issues The yearly financial losses incurred by several important industries both before and after the application of AI technology are shown in the Table 2 . For instance, the energy industry loses $ 500 million a year without AI, but when AI is implemented, those losses are cut in half to $ 150 million. The transportation, healthcare, telecommunications, financial services, and manufacturing industries all saw cutbacks of 70%. The water supply industry has experienced a somewhat lesser 60% decline, going from $ 200 million to $ 80 million each year. Table 2 Loss Reduction with AI and without AI Sector Loss Without AI ( $ Million/year) Loss With AI ( $ Million/year) Loss Reduction (%) Energy 500 150 70 Transportation 400 120 70 Healthcare 300 90 70 Water Supply 200 80 60 Telecommunications 350 105 70 Financial Services 600 180 70 Manufacturing 250 75 70 Taxonomy of AI for CI This section presents a taxonomy highlighting current issues and security challenges in CI. The proposed taxonomy categorizes AI-based security measures developed for CI and emphasizes the various tactics and solutions to lower risks. Figure 7 displays the distribution of publication categories in the Critical Infrastructure field. The data shows that a significant portion of the literature comprises conference papers (34.9%) and articles (49.6%), suggesting the continuous research and discussion occurring in academic and professional contexts. Reviews, editorials, and book chapters are among the other publications that contribute smaller but equally important parts to the body of knowledge. This distribution emphasizes how crucial peer-reviewed publications and conference talks are to advancing critical infrastructure research. For instance, (Kong et al., 2021 ) illustrate the concept of precision farming, improving agricultural productivity, and effectively managing the sustainability supply chain. They have proposed a model-based cross-stage partial network, CSPNet, three parallel sub-networks, and a cross-level fusion module. The model can classify the crop type correctly and see different crop types. The recognition accuracy, as well as the F1 score model, performs up to competitive standards. In addition, the authors have noted that implementation of the model can also be done in a lightweight manner with parameters that will enhance speed without compromising accuracy, further enhancing the system's usage. (Liu et al., 2022 ) outline development impediments in the agricultural industry. CSD shows minimal variation in the severity of the same crop disease, making it even more complicated as features of lesions are only a troublesome part of fashioning it. The manuscript presents a modified lightweight CNN to perform better in the fine-grained classification of agricultural diseases. (Broby et al., 2016) presented some cybersecurity challenges using AI in the financial sectors. Problems with ransomware attacks on financial corporations are discussed. (Gramegna et al., 2020) have proposed the XAI model with Extreme Gradient Boosting, where the users can have insight into what insurance they are buying. Fekete, A. ( 2019 ) discuss how risk reduction can be used in the trend prediction of stocks. They proposed different types of AI models: decision tree, RF, adaptive boosting (AdaBoost), support vector classifier (SVC), naïve Bayes, logistic regression, ANN, and DL methods. (Shyam et al., 2017) presented a study on the country's renewable energy and its implications on the nation's energy security and economic stability. Past scenarios related to performance of renewable resources, the present system, and the way to utilize energy in India were also discussed. (Kumar et al., 2020 ) proposed the methodology for building an intelligent healthcare system using blockchain 4.0 technology along with programming, tools, interoperability, and techniques, considering approaches of simulation and implementation for validation. (Nazar et al., 2021 ) discussed several AI models and their various uses in the health field while researching computer technology and human interaction; it is known as human-computer interaction. This study infers that XAI can be related to AI and HCI, and further exploring XAI in the healthcare field can be rewarding in terms of future direction. Table 3 presents various case studies where deep learning approaches have been applied to enhance the resilience of infrastructure systems. Each case study demonstrates how different AI techniques have been implemented across various sectors, yielding significant improvements in system stability, safety, and reliability. Table 3 Case Studies on Deep Learning in Resilient Infrastructure Systems Case Study Location Infrastructure System Deep Learning Approach Impact Smart Grid Management United States Energy CNNs, RNNs Enhanced grid stability and reduced blackout risks Intelligent Traffic Systems Singapore Transportation Reinforcement Learning Reduced traffic congestion and improved safety Smart Water Monitoring Netherlands Water Supply Autoencoders Early detection of leaks and contamination Cybersecurity in Telecom Networks South Korea Communication GANs Improved detection and mitigation of cyber threats Predictive Maintenance in Hospitals Germany Healthcare RNNs Increased equipment uptime and reliability Table 4 lists hardware solutions optimized for energy-efficient deep learning applications in infrastructure systems. Each hardware type is described along with its energy efficiency features and relevant use case examples. Table 4 Energy-Efficient Hardware for Deep Learning in Infrastructure Systems Hardware Description Energy Efficiency Features Use Case Example Tensor Processing Units (TPUs) Custom-built for neural networks High throughput with low power consumption Smart grid analytics Field-Programmable Gate Arrays (FPGAs) Reconfigurable hardware Optimized for specific tasks, leading to energy savings Real-time traffic management Graphics Processing Units (GPUs) Parallel processing capabilities Efficient for training large models with less energy per computation Predictive maintenance Application-Specific Integrated Circuits (ASICs) Customized for deep learning tasks Highly efficient in specific applications Water quality monitoring Edge Devices Devices capable of on-site processing Reduces data transmission needs and associated energy costs Real-time intrusion detection Table 5 provides examples of energy-efficient techniques applied to deep learning within various infrastructure systems, demonstrating how these techniques contribute to reduced energy consumption and enhanced system performance. Table 5 Case Studies on Energy-Efficient Deep Learning in Infrastructure Systems Case Study Location Infrastructure System Energy-Efficient Technique Impact Smart Grid Optimization United States Energy Edge Computing, TPUs Reduced operational energy costs and improved grid stability Intelligent Traffic Management Singapore Transportation Reinforcement Learning, FPGAs Decreased energy consumption and reduced traffic congestion Efficient Water Monitoring Netherlands Water Supply Quantization, ASICs Lower energy usage in continuous monitoring systems Real-Time Network Security South Korea Communication Knowledge Distillation, Edge Devices Enhanced security with reduced energy footprint Predictive Healthcare Maintenance Germany Healthcare Model Pruning, GPUs Increased equipment uptime with lower energy consumption Table 6 presents case studies on the application of deep learning for anomaly detection in various critical infrastructure systems, focusing on the performance metrics used and the impact of these approaches. Table 6 Anomaly Detection in Critical Infrastructure Systems Case Study Location Infrastructure System Deep Learning Approach Performance Metrics Impact Smart Grid Fault Detection United States Energy CNNs, RNNs Accuracy, precision, recall Improved reliability and reduced downtime Traffic Anomaly Detection Singapore Transportation LSTM Networks Mean time to detection, accuracy Enhanced traffic flow and reduced congestion Water Quality Anomaly Detection Netherlands Water Supply Autoencoders Detection rate, false negative rate Early contamination detection and improved water safety Network Security Monitoring South Korea Communication GANs, Autoencoders Detection rate, false positive rate Enhanced detection of network intrusions Healthcare Equipment Monitoring Germany Healthcare RNNs, LSTMs Sensitivity, specificity Increased equipment uptime and reliability Table 7 outlines various data integration techniques used in critical infrastructure systems, highlighting their advantages in ensuring seamless data exchange and integration. Table 7 Data Integration Techniques for Critical Infrastructure Systems Technique Description Advantages of Data Integration API Integration Connecting different systems using application programming interfaces (APIs) Enables seamless data exchange between systems ETL (Extract, Transform, Load) Extracting data from multiple sources, transforming it, and loading it into a target system Standardizes data formats and ensures consistency Data Fusion Combining data from diverse sources to produce a unified representation Enhances data completeness and accuracy Federated Learning Training machine learning models across multiple decentralized devices or servers Preserves data privacy while leveraging distributed data Ontology Alignment Aligning semantic representations of data from different sources Facilitates interoperability and knowledge sharing 3.1 Key Deep Learning Techniques Used in Critical Infrastructure Systems This section, provides an overview of deep learning architecture introduced to improve the assessment of critical infrastructure in smart cities. Figure 8 elaborates the internal flow of the entire process starting from data acquisition and preprocessing, referred to as deep learning model building, identifying the levels of feature engineering, CNN component, LSTM component, fully connected network, reaching real time criticality assessment. The first instance in the process is the gathering of data from different places containing information related to Flooding Level, Total Electricity Use, Total Energy Consumption, Population, and Poverty Percent. This data which has so far been collected is subjected to other processing activities which include missing value imputes, one hot encoding of fields with categories and normalizing the values. Then comes the stage of feature engineering in which new variables such as Population to Area Ratio (Pdensity) and Energy Per Capita (Ecapita) are designed in order to make the model better at predicting. 3.1.1 Convolutional Neural Networks (CNNs) CNNs are widely employed in many similar settings, such as security cameras or drone photography, to monitor the condition of structures, such as buildings, bridges, and highways. Smart modern grid systems make use of their skills to detect defects in transformers and electrical wires for better maintenance scheduling and grid stability. According to a study, CNN predicts malfunctions in power grid components and develops a solution before they cause service interruptions. CNNs can analyze satellite photos or grid layouts to detect changes or failures in the physical infrastructure of pipelines, electrical grids, or transportation networks. CNNs are able to analyze video streams in critical infrastructure or transportation settings (such airports and water treatment plants) in order to spot anomalies, malfunctions, or security breaches. CNNs can recognize abnormalities in smart grid sensor networks and notify users of potential infrastructure problems because they are trained on the spatial patterns of normal operating conditions. 3.1.2 Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) Networks RNNs, most importantly the Long Short Term Memory networks (LSTM), are useful and bring substantial benefits to models that need to do predictive analytics on sequential data, like time series data in CIS. They, in line with the case, can also encode time-based information and trends in the sensor data, which is paramount to observing how the system behaves over time. In the load dispatching and resource optimization in the power distribution network, LSTM's forecasts the demand on the grid and the patterns of traffic flow control systems. For instance, energy prediction through the application of LSTM in intelligent grids has helped achieve lower energy costs and improved energy management. LSTMs can predict when critical mechanical apparatus (like turbines, generators, and train engines), based on time series variables such as temperature and vibration, may fail by predicting days on which failure is likely to occur. The short-term and long-term electrical demand may be projected in smart grids using LSTM networks. Although historical factors generally serve as a foundation, forecasts, and current factors are also incorporated. In industrial control systems (ICS), LSTMs detect security breaches by keeping track of traffic data or system logs to identify irregularities that cyber-attacks or ineffective operations can cause. 3.1.3 Autoencoders Major interdependent application areas of the manor are anomaly detection based on sensor data and location issues such as cyber-attacks or other similar problems in the computer industry. They help in different areas, especially cyberspace, where many networks are monitored, and invisible network patterns are searched. Autoencoders can capture the general functioning behavior of important systems and screen for out-of-band events or anomalies that jeopardize those systems. Autoencoders can also reduce sensor data for sending over moderately constrained areas such as offshore wind farms. Systems like autoencoders are used to detect changing levels of operation that are usually associated with a decline in the performance of specific equipment or the presence of faults. For example, autoencoders were used to detect industrial control system anomalies to prevent unplanned outages. 3.1.4 Generative Adversarial Networks (GANs) Synthetic data generation is performed using GANs in applications where creating rigorous data is difficult for training DL models. This method finds application in scenario modeling to simulate complex attacks on critical infrastructure. The advantages of GANs are that artificial data can be used to improve the capability of further deploying anomaly detection models in smart grids. The most typical hostile applications of GANs are generating adversarial examples to test ICS cyber security more convincingly than usual. GANs help create fake sensor data to teach models that are intended for detecting faults or predicting the maintenance of industrial machines, especially When there is a lack of sufficient real data. There is a simulation of extreme conditions, which propagates overloading of the Power grid, and managers of the CIS can enhance their capabilities. 3.1.5 Reinforcement Learning (RL) RL is employed in complex autonomous systems like drones and self-driving cars for instant decisions. It is also possible to vary the electricity distribution to fit the existing supply and consumption patterns, making smart grid systems economically beneficial. For example, RL has been reported to solve the traffic optimization problem where traffic signals were adjusted optimally to alleviate traffic congestion through RL. Similarly, smart grids are implemented using Deep RL algorithms, which manage optimal demand and supply, storage, and reduce costs in a volatile market. In smart cities, however, using distributed reinforcement learning (DRL) there is a possibility of accomplishing instant traffic control, manipulation of signals so that stagnation is avoided, or dispersal of pathways in cases of terrorism. When looking at the cloud infrastructure, DLR can also be used in data centers where resource allocation is carried out, and energy usage is optimized during processing. 3.1.6 Graph Neural Networks (GNNs) In the case of power systems and water distribution systems, GNNs also assess and optimize networked infrastructure systems. They are used for fault isolation considering the interconnections of network elements. GNNs were applied in the design of water distribution systems to enhance network resilience and mitigate the consequences of water pipeline failures. In comparison, GNNs provide an effective means of encapsulating the complexities of a power grid system by treating it as one large network consisting of nodes, ie substation and transmission lines, and edges, i.e power lines. This is useful for fault detection, cascade failure prediction, and enhancement of grid management. GNNs for intrusion detection systems (IDS) are specialized to identify abnormal behavior or attacks over any communications applications that provide the essence of the critical infrastructure systems. GNNs optimize traffic circulation across intersections by modeling inter intersection flows and router distributions. 3.2 Factors Influencing Infrastructure Resilience Maintaining and enhancing infrastructure systems is challenging because of the complex interactions between many factors. Recent studies examining various elements that affect infrastructure resilience (IR) have highlighted the strengths and weaknesses of the current infrastructure systems (Almaleh et al., 2024). 3.2.1 Environmental Factors: Natural disasters, including hurricanes, cyclones, and floods, have a significant effect on infrastructure systems, especially in Small Island Developing States (SIDS) (Henriques et al., 2023 ). These circumstances seriously jeopardize the resilience and recoverability of the energy infrastructure. For example, it was shown that the fragility of the systems and the recovery time after damage account for a large portion of the low resilience of the SIDS energy infrastructure. The report also showed how government actions could drastically reduce these risks by strengthening infrastructure resilience and accelerating recovery times. 3.2.2 Engineering and Design: A critical factor in infrastructure's capacity to absorb shocks and bounce back is its physical robustness. Their engineering and design primarily determine the robustness of these systems. One of the most critical factors influencing system performance is the infrastructure's internal dynamics, including interactions between its physical components, human operators, and external stresses (Chen et al., 2023 ). Their study demonstrated how these internal mechanisms, when coupled with ongoing stressors, may lessen the efficacy of infrastructure and increase its susceptibility to malfunctions. 3.2.3 Human Factors: People who work in infrastructure systems—managers, operators, and policymakers—significantly impact their resilience. How these actors interact with the physical infrastructure could enhance or impair the system's capacity to adjust to and bounce back from changes. Resilience requires efficient management and timely responses by human actors, especially when confronted with acute shocks. 3.2.4 Chronic and Acute Stressors: Short- and long-term pressures can reduce the resilience of infrastructure systems. Over time, system performance is steadily reduced by chronic stresses such as aging infrastructure, increased demand, and decreasing environmental conditions (Yang et al., 2023 ). However, acute stressors—like industrial mishaps, cyberattacks, or natural disasters—bring severe and unexpected difficulties. One of the most critical aspects of infrastructure is its capacity to bounce back quickly from these disastrous occurrences. 3.2.5 Policy and Governance: Infrastructure resilience is significantly impacted by policy decisions. Effective rules can ensure that resources are available for upgrades and maintenance, shorten response and recovery times, and support robust infrastructure planning and execution. Targeted policy initiatives could strengthen the overall resilience of the energy infrastructure in these nations by addressing the unique vulnerabilities in the SIDS setting. 3.2.6 Simulation and Modeling: Using models, like agent simulations, has helped assess how resilient infrastructure systems are in various scenarios. In 2022, a simulation model was created to mirror the interactions between infrastructure, real people, and stress factors. This model provides insights into how these factors affect the long-term resilience of infrastructure systems (Luiijf et al., 2021). To identify vulnerabilities and develop strategies to enhance resilience, it's essential to simulate various scenarios and challenges. 3.3 Case Studies 3.3.1 AI for Predictive Maintenance and Grid Optimization in Smart Grids The application of Artificial Intelligence (AI) in smart grids has proven transformative in enhancing the resilience, efficiency, and reliability of energy infrastructure. A European power utility implemented an AI-driven predictive maintenance and grid optimization system to address challenges such as equipment failures, inefficient energy distribution, and the integration of renewable energy. IoT sensors deployed across transformers and transmission lines continuously monitored operational parameters like voltage and temperature, providing real-time data for AI-based analysis. Machine learning models, including Random Forest and LSTM networks, were used to predict potential equipment failures, allowing for proactive maintenance and reducing unscheduled downtimes by 40%. Additionally, the AI system optimized load balancing and energy flow, improving overall energy efficiency by 20% and enabling the integration of 30% more renewable energy. The predictive capabilities of the AI system extended the lifespan of critical components by 25% while reducing operational costs by 15%. AI-enhanced cybersecurity measures also safeguarded the grid from potential cyber threats, further ensuring the stability of the power network. This case study demonstrates how AI technologies can significantly improve the operational performance and sustainability of smart grids, making them a vital component of modern energy infrastructure. An overview of significant smart grid initiatives from different nations is given in the Table 8 , which also includes information on the projects' costs, technology employed, start year, descriptions, and results. Large-scale smart grid technology was pioneered by projects like Enel's Telegestore in Italy, which began operations in 2005 with an investment of €2.1 billion and generated yearly savings of €500 million. One of the biggest grid modernization projects is the US Department of Energy's ARRA Smart Grid Project (2009), which invested over $ 9 billion in technology like smart meters, cybersecurity, and renewable energy integration. Other noteworthy projects include Hydro One's program in Ontario, which was praised for its wide coverage and cutting-edge metering technology, and Chattanooga's smart grid in Tennessee, which reduced power outages by 60% and saved $ 60 million yearly. Table 8 Global Smart Grid Projects Project Description Year Cost Technologies Outcome Enel - Telegestore The Italian Telegestore project by Enel S.p.A. was one of the earliest and largest smart grid implementations. Enel designed and manufactured its own meters and software, providing commercial-scale smart grid technology to homes. 2005 €2.1 billion Smart meters, system integration, custom software development Annual savings of €500 million; pioneer in smart grid technology deployment. US Dept. of Energy - ARRA Smart Grid Project A large-scale initiative funded by the American Recovery and Reinvestment Act (ARRA) of 2009, involving over $ 9 billion in public and private funds. It included technologies such as advanced metering infrastructure, grid automation, and cybersecurity projects. 2009 $ 9 billion+ Advanced Metering Infrastructure, smart meters, synchrophasors, energy storage, cybersecurity, renewable energy integration Enhanced grid efficiency; reports on impact due by Q2 2015. Austin, Texas Austin’s smart grid project began in 2003 with the replacement of manual meters by smart meters. The grid supports real-time management of smart devices and meters across the city. 2003 N/A Wireless mesh network, smart meters, smart thermostats, sensors Managed 200,000 devices, aiming for 500,000; supports 1 million consumers and 43,000 businesses. Boulder, Colorado Boulder completed the first phase of its smart grid project in 2008. The system uses smart meters as gateways to home automation networks (HANs) controlling various smart devices. 2008 N/A Smart meters, home automation networks (HANs) Initial phase completed; integrates home automation with smart metering. Hydro One Hydro One’s smart grid initiative in Ontario, Canada involves deploying a standards-compliant communication infrastructure. The project aims to serve 1.3 million customers. 2010 N/A Standards-compliant communication infrastructure, advanced metering Awarded "Best AMR Initiative in North America"; extensive coverage. Sydney Sydney implemented the Smart Grid, Smart City program in partnership with the Australian Government. N/A N/A Smart grid technologies, city-wide deployment N/A Évora - InovGrid The InovGrid project in Évora, Portugal focuses on automating grid management and improving service quality. It includes advanced grid control and renewable energy integration. N/A N/A Grid automation, renewable energy integration, electric vehicle management Advances in grid management, service quality, and energy efficiency. Massachusetts Proposed smart grid project rejected due to concerns about impacts on low-income customers. 2009 N/A Smart meters, pre-pay billing, dynamic rates Rejected due to concerns about fairness and protections for low-income customers. Chattanooga Smart grid with power-line interrupters and fiber-optic systems; includes gigabit-speed internet. 2008 $ 232.2 million Power-line interrupters, smart meters, fiber-optics Reduced outages by 60%, saved $ 60 million annually. China Developing smart grid with demand response pilot using OpenADR standard. N/A $ 22.3 billion (market size) Demand response, OpenADR standard Significant growth projected in smart grid market; ongoing pilot studies. United States DOE-funded demand response programs using OpenADR standard. 2009 $ 22.4 million Demand response, OpenADR standard Implemented demand response programs. Hawaiian Electric Co. (HECO) Testing ADR program to manage wind power intermittency in Hawaii. N/A N/A ADR program, renewable energy integration Pilot for managing renewable energy. 3.3.2 AI-Driven Predictive Analytics in ICU for Enhanced Patient Monitoring and Care The integration of Artificial Intelligence (AI) in healthcare has markedly improved patient care and operational efficiency, as demonstrated by a prominent urban hospital's implementation of an AI-powered predictive analytics system in its Intensive Care Unit (ICU). This system utilized Internet of Medical Things (IoMT) devices to continuously monitor vital signs such as heart rate, blood pressure, and oxygen levels, transmitting real-time data to a centralized analytics platform. Employing deep learning models, particularly LSTM networks, the system analyzed these data streams to predict critical health events like sepsis and cardiac arrest before they occurred. This proactive approach allowed for timely interventions, reducing ICU response times by 30% and decreasing patient mortality rates by 15%. The AI system also optimized ICU resource utilization by identifying patients who could be safely monitored outside the ICU, leading to a 20% reduction in the average length of ICU stays and significant cost savings. Additionally, the system’s predictive capabilities enhanced staff decision-making by providing actionable insights and real-time alerts, thus transforming the management of critical care and improving overall patient outcomes. This case study highlights the profound impact of AI in revolutionizing healthcare delivery through early detection, efficient resource management, and enhanced clinical decision support. Challenges and Future Opportunities Incorporating AI, specifically DL, into critical infrastructure systems has improved their resilience and efficiency. However, there are several issues with this relationship. Figure 9 illustrates the challenges involved in developing essential infrastructure systems. 1. AI Model Complexity and Overfitting Overfitting and abnormal behaviors can be challenging to manage in AI models due to their complexity, particularly in those that use DL techniques. While deep learning models produce robust predictions, their black-box structure and overfitting risk need the development of methods to reduce computational overhead and improve generalizability (Raval et al., 2023 ). 2. Scalability Many challenges arise when managing large volumes of data in real time (Wang et al., 2023 ). Scalable systems must balance security and performance since increasing complexity may lead to vulnerabilities. Building dependable systems that can handle complex data requirements while maintaining security is essential. 3. Logic Transparency The black-box nature of AI models hinders explainability. When decision-making requires transparency, deep learning models' interpretability could provide difficulties. Research on explainable AI (XAI) is critical to enhancing understanding and confidence in AI-generated judgments (Dosso et al., 2023; Tiong et al., 2023). 4. Resilience and Adaptability Resilience is a system's ability to withstand shocks and recover, whereas adaptability is a system's ability to adjust to new situations (Luiijf et al., 2021). AI systems must be adaptable to new situations and capable of handling evolving threats to be resilient. 4.1 Future Opportunities Quantification and Reduction of Uncertainty: ML can address the uncertainties in post-disruption recovery by combining data from various sources to quantify uncertainty and improve stochastic optimization models (Ampratwum et al., 2022 ). Understanding Dependencies: Integrating ML with historical disruption databases can enhance understanding of infrastructure dependencies, improving prevention, mitigation, and recovery strategies (Vargas et al., 2023; Alkhaleel et al., 2023). Human Factor Integration: ML can simulate the impact of human characteristics on infrastructure systems, identifying potential failure points and improving resilience through better decision-making (Hassanzadeh et al., 2020 ). Dynamic and Risk-Averse Optimization: RL techniques can introduce flexibility into resilience models, allowing for sequential adjustments and risk-averse decision-making under high-risk conditions. Wearable IoT Integration: Wearable devices to gather data on community movement, health, and environmental conditions can enhance data-driven decision-making in disaster scenarios, though they must address privacy and security concerns (Jadav, 2023; Roopashree, 2022; Vargas, 2023 ; Xie, 2019). 4.2 Data Sources To guarantee a thorough study, the data used for this assessment was painstakingly collected from various digital and electronic sources. Prominent digital libraries like IEEE Xplore, ACM Digital Library, ScienceDirect, Elsevier, Springer, and Wiley were among our primary sources. We also referred to various technical literature sources, such as industry reports, books, peer-reviewed academic articles, technical blogs, and patents. Our ability to gather a wide range of sources allowed us to create a solid dataset for the review. 4.3 Search Techniques One of the procedures involved using specific search terms related to "critical infrastructure," along with terms like "machine learning," "artificial intelligence," "smart grid," and "cybersecurity." These keywords were explicitly selected to encompass a wide range of relevant papers from various academic disciplines. A search strategy was developed to ensure that the most recent and pertinent documents were included in the review. 4.4 Data Availability and Quality Creating AI models for cybersecurity in critical infrastructure requires access to substantial amounts of high-quality data (Luiijf et al., 2021). Such data is necessary for training models to correctly distinguish between benign and malicious behaviors or attack and non-attack scenarios. However, open-source repositories need more relevant datasets (Alkhaleel et al., 2023). The existing datasets often contain outdated feature spaces for training models specific to contemporary security systems. High-quality, up-to-date data is crucial for developing AI models that provide accurate and reliable classification (M. Lindström et al., 2009; Zhou et al., 2020 ). The effectiveness of cybersecurity measures for vital infrastructure could be diminished if old or insufficient datasets are used, as this could lead to erroneous classifications. 4.5 Funding sponsors The influence of funding sponsors on research productivity is critical to understanding how scientific innovation is fostered globally. Figure 10 shows the contributions of significant funding sponsors to enhance academic research on CI. The analysis focuses on data from prominent international sponsors, including the National Science Foundation, the National Natural Science Foundation of China, and the European Commission. Conclusion The discussion in this paper was related to AI implementation in cybersecurity, focusing on critical infrastructure, the transformative potential of AI-based solutions, and challenges regarding the protection of such systems. We provided a comprehensive literature review, historical analysis, and an in-depth case study underlined AI's massive impact on detecting, preventing, and mitigating cyber threats. The study also focused on undecided issues like data privacy, algorithmic bias, and changing human operator roles in the AI-driven security environment. Our findings underline the need for responsible and ethical application of AI in effectively safeguarding critical infrastructure. With ongoing research and collaboration, robust development of AI systems becomes indispensable in assuring resilience and security for critical infrastructure against emerging cyber threats. Declarations Author Contribution GHL wrote the main manuscript and reviewd the manuscript Acknowledgments There is no conflict of interest References Jadav NK, Gupta R, Tanwar S (2023), January AI and onion routing-based secure architectural framework for IoT-based critical infrastructure. In 2023 13th International Conference on Cloud Computing, Data Science & Engineering (Confluence) (pp. 559–564). IEEE Adil M, Khan MK, Jadoon MM, Attique M, Song H, Farouk A (2022) An AI-enabled hybrid lightweight authentication scheme for intelligent IoMT based cyber-physical systems. IEEE Trans Netw Sci Eng 10(5):2719–2730 Alyahya S, Khan WU, Ahmed S, Marwat SNK, Habib S (2022) Cyber secure framework for smart agriculture: Robust and tamper-resistant authentication scheme for IoT devices. Electronics 11(6):963 Annadurai C, Nelson I, Devi KN, Manikandan R, Jhanjhi NZ, Masud M, Sheikh A (2022) Biometric authentication-based intrusion detection using artificial intelligence internet of things in smart city. Energies 15(19):7430 Garabato D, Dafonte C, Santovena R, Silvelo A, Novoa FJ, Manteiga M (2022) AI-based user authentication reinforcement by continuous extraction of behavioral interaction features. Neural Comput Appl 34(14):11691–11705 Lee K, Yim K (2020) Cybersecurity threats based on machine learning-based offensive technique for password authentication. Appl Sci 10(4):1286 Roopashree S, Anitha J, Mahesh TR, Kumar VV, Viriyasitavat W, Kaur A (2022) An IoT based authentication system for therapeutic herbs measured by local descriptors using machine learning approach. Measurement 200:111484 Met I, Erkoç A, Seker SE (2022) Performance, efficiency, and target setting for bank branches: time series with automated machine learning. IEEE Access 11:1000–1010 Nabipour M, Nayyeri P, Jabani H, Shahab S, Mosavi A (2020) Predicting stock market trends using machine learning and deep learning algorithms via continuous and binary data; a comparative analysis. Ieee Access 8:150199–150212 Kochunas B, Huan X (2021) Digital twin concepts with uncertainty for nuclear power applications. Energies 14(14):4235 Blessy JA (2021), February Smart irrigation system techniques using artificial intelligence and IoT. In 2021 Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV) (pp. 1355–1359). IEEE Hemming S, de Zwart F, Elings A, Righini I, Petropoulou A (2019) Remote control of greenhouse vegetable production with artificial intelligence—greenhouse climate, irrigation, and crop production. Sensors 19(8):1807 Xu L, Zhou X, Tao Y, Liu L, Yu X, Kumar N (2021) Intelligent security performance prediction for IoT-enabled healthcare networks using an improved CNN. IEEE Trans Industr Inf 18(3):2063–2074 Wazid M, Das AK, Mohd N, Park Y (2022) Healthcare 5.0 security framework: applications, issues and future research directions. IEEE Access 10:129429–129442 Nazar M, Alam MM, Yafi E, Su’ud MM (2021) A systematic review of human–computer interaction and explainable artificial intelligence in healthcare with artificial intelligence techniques. IEEE Access 9:153316–153348 Ale L, Sheta A, Li L, Wang Y, Zhang N (2019), December Deep learning based plant disease detection for smart agriculture. In 2019 IEEE Globecom Workshops (GC Wkshps) (pp. 1–6). IEEE Dhieb N, Ghazzai H, Besbes H, Massoud Y (2020) A secure ai-driven architecture for automated insurance systems: Fraud detection and risk measurement. IEEE Access 8:58546–58558 Kong J, Wang H, Wang X, Jin X, Fang X, Lin S (2021) Multi-stream hybrid architecture based on cross-level fusion strategy for fine-grained crop species recognition in precision agriculture. Comput Electron Agric 185:106134 Liu Y, Gao G, Zhang Z (2022) Crop disease recognition based on modified light-weight CNN with attention mechanism. IEEE Access 10:112066–112075 Broby D, Karkkainen T (2016) FINTECH in Scotland: building a digital future for the financial sector. The Future of Fintech Supported by International Financial Services District (IFSD) The Technology Innovation Centre, Glasgow Date: 2nd September Gramegna A, Giudici P (2020) Why to buy insurance? An explainable artificial intelligence approach. Risks 8(4):137 Fekete A (2019) Critical infrastructure and flood resilience: Cascading effects beyond water. Wiley Interdisciplinary Reviews: Water, 6(5), e1370 Shyam B, Kanakasabapathy P (2017), December Renewable energy utilization in India—policies, opportunities and challenges. In 2017 International Conference on Technological Advancements in Power and Energy (TAP Energy) (pp. 1–6). IEEE Kumar A, Krishnamurthi R, Nayyar A, Sharma K, Grover V, Hossain E (2020) A novel smart healthcare design, simulation, and implementation using healthcare 4.0 processes. IEEE access 8:118433–118471 Nazar M, Alam MM, Yafi E, Su’ud MM (2021) A systematic review of human–computer interaction and explainable artificial intelligence in healthcare with artificial intelligence techniques. IEEE Access 9:153316–153348 Almaleh A (2024) A Novel Deep Learning Approach for Real-Time Critical Assessment in Smart Urban Infrastructure Systems. Electronics 13(16):3286 Henriques J, Caldeira F, Cruz T, Simões P (2023) A forensics and compliance auditing framework for critical infrastructure protection. Int J Crit Infrastruct Prot 42:100613 Chen J, Lu Y, Zhang Y, Huang F, Qin J (2023) A management knowledge graph approach for critical infrastructure protection: Ontology design, information extraction and relation prediction. Int J Crit Infrastruct Prot 43:100634 Yang Z, Barroca B, Laffréchine K, Weppe A, Bony-Dandrieux A, Daclin N (2023) A multi-criteria framework for critical infrastructure systems resilience. Int J Crit Infrastruct Prot 42:100616 Luiijf E, Klaver M (2021) Analysis and lessons identified on critical infrastructures and dependencies from an empirical data set. Int J Crit Infrastruct Prot 35:100471 Raval KJ, Jadav NK, Rathod T, Tanwar S, Vimal V, Yamsani N (2023) A survey on safeguarding critical infrastructures: Attacks, AI security, and future directions. Int J Crit Infrastruct Prot, 100647 Wang S, Sun J, Zhang J, Dong Q, Gu X, Chen C (2023) Attack-Defense game analysis of critical infrastructure network based on Cournot model with fixed operating nodes. Int J Crit Infrastruct Prot 40:100583 Dosso YS, Rizcallah E, Kwamena F, Goubran R, Green JR (2022), December Deep Learning for Segmentation of Critical Electrical Infrastructure from Vehicle-Based Images. In 2022 IEEE Electrical Power and Energy Conference (EPEC) (pp. 241–247). IEEE Tiong A, Vergara HA (2023) Evaluation of network expansion decisions for resilient interdependent critical infrastructures with different topologies. Int J Crit Infrastruct Prot 42:100623 Ampratwum G, Osei-Kyei R, Tam VW (2022) Exploring the concept of public-private partnership in building critical infrastructure resilience against unexpected events: A systematic review. Int J Crit Infrastruct Prot 39:100556 Vargas P, Tien I (2023) Impacts of 5G on cyber-physical risks for interdependent connected smart critical infrastructure systems. Int J Crit Infrastruct Prot 42:100617 Alkhaleel BA (2023) Machine learning applications in the resilience of interdependent critical infrastructure systems—A systematic literature review. Int J Crit Infrastruct Prot, 100646 Papadopoulos L, Demestichas K, Muñoz-Navarro E, Hernández-Montesinos JJ, Paul S, Museux N, Levak J (2024) Protection of critical infrastructures from advanced combined cyber and physical threats: The PRAETORIAN approach. Int J Crit Infrastruct Prot 44:100657 Lindström M, Olsson S (2009) The European program for critical infrastructure protection. Crisis Management in the European Union: Cooperation in the Face of Emergencies. Springer, pp 37–59 Zhou S, Yang Y, Ng ST, Xu JF, Li D (2020) Integrating data-driven and physics-based approaches to characterize failures of interdependent infrastructures. Int J Crit Infrastruct Prot 31:100391 Hassanzadeh A, Rasekh A, Galelli S, Aghashahi M, Taormina R, Ostfeld A (2020) Katherine Banks. A review of cybersecurity incidents in the water sector. J Environ Eng 146(5):03120003 Xie J, Stefanov A, Liu CC (2019) Physical and cybersecurity in a smart grid environment. Adv Energy Systems: Large-scale Renew Energy Integr Chall, 85–109 Additional Declarations No competing interests reported. Supplementary Files floatimage1.png Graphical/Visual Abstract and Caption Critical Infrastructure Integration with Deep Learning for Enhanced Infrastructure Resilience Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5923379","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":408637448,"identity":"b87796ed-1d34-48f6-a327-723aed798a61","order_by":0,"name":"H L Gururaj","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/klEQVRIiWNgGAWjYPACNgYDduYDQIYEAwM7kOIhSgszWwJECzNxWhiAWngMICxCWuQjcp9JV9TwMZgz83zd8HOHRR5/MwPjg7dtuLUY3kg3kzxzjI3Bspl3283eMxLFEocZmA3n4tMyI41NsoEN6JfDvNtu8LZJJDYcZmCT5iWo5R9IC8+zm3+BWuYfZmD/jU+LvARQS2MbWAvbbZAtG4C2MOPTYsDzjNmysY+Nx+Awm9lt2TaJYsPDjM2Sc87hsaU9jfFmw7djcgbHm5/dfNtWlyd3vPnghzdleGw5AKaOwSMigYGBsQG3epAtEOkauEACXuWjYBSMglEwIgEAiI1KGSl0afYAAAAASUVORK5CYII=","orcid":"","institution":"Manipal Academy of Higher Education","correspondingAuthor":true,"prefix":"","firstName":"H","middleName":"L","lastName":"Gururaj","suffix":""}],"badges":[],"createdAt":"2025-01-29 10:08:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5923379/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5923379/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":75174406,"identity":"6b043044-9d3e-4de6-bb14-53705a339bfb","added_by":"auto","created_at":"2025-01-31 14:59:18","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":93072,"visible":true,"origin":"","legend":"\u003cp\u003eCyber Attacks\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/75e8d7d568a50fe16e288d19.jpeg"},{"id":75174770,"identity":"05e43adb-5762-469b-b46c-33ef79f478e2","added_by":"auto","created_at":"2025-01-31 15:07:18","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":39852,"visible":true,"origin":"","legend":"\u003cp\u003eEstimated Financial Losses Due to Different Cyber Attacks\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/dc60ae94886afd2b0d750da1.jpeg"},{"id":75175586,"identity":"abfac22d-0820-4bcc-8146-10064a70fec6","added_by":"auto","created_at":"2025-01-31 15:15:18","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":44747,"visible":true,"origin":"","legend":"\u003cp\u003eStatistics of cybercrimes in critical infrastructures\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/c73058ebd7ef8b47b7d9d549.png"},{"id":75174772,"identity":"bf0d6445-4984-4a7d-9b92-81b26427f447","added_by":"auto","created_at":"2025-01-31 15:07:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":32566,"visible":true,"origin":"","legend":"\u003cp\u003ePublication by year [Source from Scopus]\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/96871a4cef373266813f4eb8.png"},{"id":75174402,"identity":"05d89533-605c-44ad-b689-f24cb475dd0c","added_by":"auto","created_at":"2025-01-31 14:59:18","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":30123,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of Sources\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/f7f6a325f1d67516841b345b.png"},{"id":75174426,"identity":"a041147d-3161-45fc-8ec5-5a262dd49661","added_by":"auto","created_at":"2025-01-31 14:59:18","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":245126,"visible":true,"origin":"","legend":"\u003cp\u003eCategories of CIS (Critical Infrastructure Systems)\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/a7bd5f84217c15f9179e89ac.png"},{"id":75174412,"identity":"5d6ac4cb-a01e-4f9e-9154-5f4ddf882f9e","added_by":"auto","created_at":"2025-01-31 14:59:18","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":60651,"visible":true,"origin":"","legend":"\u003cp\u003ePublication type [Sourced from Scopus]\u003c/p\u003e","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/0aca3e8aab75e04272fa42bf.png"},{"id":75174773,"identity":"95694571-fc68-4b57-b55d-b0014866f11c","added_by":"auto","created_at":"2025-01-31 15:07:18","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":28914,"visible":true,"origin":"","legend":"\u003cp\u003eDeep learning model for criticality assessment\u003c/p\u003e","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/a9accef85f11c553f871b9fd.png"},{"id":75175587,"identity":"024c0689-60b8-4ab1-92ba-35563b93e30e","added_by":"auto","created_at":"2025-01-31 15:15:18","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":23593,"visible":true,"origin":"","legend":"\u003cp\u003eResearch Challenges (Broby et al. 2016)\u003c/p\u003e","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/c600961eafdc46b9be19e4db.png"},{"id":75174413,"identity":"f5b0a8eb-1b05-46bf-89b3-d503da7b997f","added_by":"auto","created_at":"2025-01-31 14:59:18","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":35042,"visible":true,"origin":"","legend":"\u003cp\u003eFunding Sponsors\u003c/p\u003e","description":"","filename":"floatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/e3146de4d7dbebb7a86aa113.png"},{"id":75606043,"identity":"dd100ddb-41f7-4548-a855-3a22325e7029","added_by":"auto","created_at":"2025-02-06 09:32:11","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2033559,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/af32a195-86d3-44c4-b080-1a58fee5145e.pdf"},{"id":75174409,"identity":"1f2f2823-52d7-4eb4-9b08-e97cc1e8dff0","added_by":"auto","created_at":"2025-01-31 14:59:18","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":147105,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGraphical/Visual Abstract and Caption\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCritical Infrastructure Integration with Deep Learning for Enhanced Infrastructure Resilience\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-5923379/v1/5c00364d99c3f111f4508de7.png"}],"financialInterests":"No competing interests reported.","formattedTitle":"Deep Learning Applications in the Resilience Critical Infrastructure Systems—A Systematic Review","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe global wave of digital transformation has opened new avenues for cyber attackers, making many individuals vulnerable to phishing and other malicious activities (Jadav et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e). From small shopkeepers using UPI applications to supermarkets relying on mobile applications, digitalization has become pervasive. Unfortunately, this widespread adoption has increased cyberattacks targeting these digital platforms.\u003c/p\u003e\n\u003cp\u003eCI is basically the backbone of economic growth and development in a nation\u0026apos;s journey towards prosperity. The arrival of IoT has brought a sea change in traditional CI, making them more connected and intelligent systems. Such intelligent transformation easily coordinates and communicates between large chunks of CI, such as power grids, transportation networks, and healthcare systems, with a small amount of data exchange. However, IoT networks are often based on obsolete legacy protocols that degrade the operational performance of CI (Adil et al.,2022; Alyahya et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e). This makes the device quite vulnerable because an adversary can easily mount a cyberattack that can range from man-in-the-middle attacks to DDoS and malware attacks.\u003c/p\u003e\n\u003cp\u003eThe integration of computer technology across various sectors has led to an increased risk of cyberattacks. These attacks have far-reaching impacts, affecting fields such as education, military, healthcare, communications, finance and banking, manufacturing industries, transportation, and software and hardware vendors, as illustrated in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\n\u003cp\u003eTechnological advancements have driven the adoption of IoT across multiple domains, enabling easier monitoring, customization, and data updating. However, identifying and eradicating cyberattacks remains a formidable challenge. Researchers have proposed cryptographic solutions, such as authentication and authorization, to ensure that only authorized individuals can access critical facilities and prevent unauthorized access attempts. These solutions often involve verifying identities through passwords or biometrics (Annadurai et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eArtificial intelligence-based systems can effectively reduce the risks associated with security threats in \u0026lsquo;Critical Infrastructure contexts despite their huge limitations (Garabato et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e; Lee et al., 2020). Several appraisal surveys or review materials illustrating the application of AI on security matters have been made over time. However, most of these surveys are irrelevant to the CI applications (Roopashree et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e). Therefore, it is essential to research and analyze the scope of activities involving managing security and privacy issues in CI employing AI. With this, we propose \u0026ndash; a systematic review that explores the state of affairs in AI in the context of cybersecurity events and issues and the prospects and barriers for AI in CI.\u003c/p\u003e\n\u003cp\u003eFigure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e displays the estimated financial losses in dollars from various types of cyberattacks. Malware is the most damaging attack, causing approximately \u003cspan\u003e$\u003c/span\u003e23.6 million in losses, followed by web-based attacks at \u003cspan\u003e$\u003c/span\u003e20.1 million. Denial of services, malicious insiders, and phishing \u0026amp; social engineering also contribute significant losses, ranging from \u003cspan\u003e$\u003c/span\u003e1.3 million to \u003cspan\u003e$\u003c/span\u003e1.6 million. Lesser but notable losses are attributed to malicious code, stolen devices, ransomware, and botnets, with botnets accounting for the most minor financial loss at \u003cspan\u003e$\u003c/span\u003e350,012. The chart highlights the need for robust cybersecurity measures to mitigate the economic impact of these prevalent threats.\u003c/p\u003e\n\u003cp\u003eFigure \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e provides a statistical overview of cybercrime incidents across various sectors. These statistics highlight the growing threat landscape and emphasize the need for robust cybersecurity measures, particularly in critical infrastructures.\u003c/p\u003e\n\u003cp\u003eWhile numerous studies exist on cybersecurity and AI, few focus specifically on AI-based cybersecurity for CI, as shown in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. Existing surveys and review articles primarily address security concerns but must include a comprehensive taxonomy highlighting AI approaches\u0026apos; intuitive nature in confronting these concerns. As Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e shows, most of the primary studies were published in ScienceDirect. Additionally, they often lack a proof-of-concept demonstrating why AI is a pivotal technology in cyberspace. Based on these observations, our proposed survey makes the following significant contributions:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\n \u003cp\u003ePresent a thorough review of AI-based Cybersecurity in CI applications.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eThis review covers critical sectors such as the financial sector (banking, insurance, and stock market), the energy industry (nuclear power plants \u0026amp; sustainable energy203), agriculture, public health, and relevant security features.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eWe offer a more complex taxonomy as it draws solutions for such critical areas using various AI techniques.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eFinally, discuss emerging scientific areas that inhibit the progress of CI and do not have a resistive preventive approach.\u003c/p\u003e\n \u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe paper is organized as follows: Section II explores various critical infrastructures, Section III examines the taxonomy of AI, Section IV discusses challenges and future opportunities, and Section V presents the conclusion.\u003c/p\u003e\n\u003cdiv id=\"Sec2\" class=\"Section2\"\u003e\n \u003ch2\u003e2.1 Critical Infrastructure\u003c/h2\u003e\n \u003cp\u003eIntegrating sensors and actuators across sectors and technologies is called Critical Infrastructure (CI). CI serves as the backbone for human existence in everyday life scenarios. The US government oversees a variety of CI within society, as illustrated in Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003e2.2 Chemical Sector\u003c/h2\u003e\n \u003cp\u003eCyberattacks against storage facilities can cause environmental disasters through sabotage or unlawful access to hazardous goods, as well as the theft of intellectual property and compromise of private data. The Chemical Sector is particularly susceptible to cyber threats due to the sensitivity of the items stored there and the potential for significant consequences on the environment and public safety. Storage facilities are vulnerable to these risks because they often handle dangerous materials, confidential data, and cutting-edge proprietary techniques.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003e2.3 Commercial Facilities Sector\u003c/h2\u003e\n \u003cp\u003eCybersecurity in the corporate facilities sector has become more concerning due to the heavy reliance on digital technology in office buildings, retail stores, and other commercial spaces. Because these institutions use networked systems and intelligent technology to improve customer service and optimize operations, they are especially vulnerable to cyberattacks that target critical infrastructure.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003e2.4 Communications Sector\u003c/h2\u003e\n \u003cp\u003eIt is evident that the telecommunication industry is vital in sustaining the world\u0026apos;s communication infrastructure, which is essential for daily living, commerce, and defense. As these networks worldwide expand and become increasingly available, this industry continues to experience an expanding number of cybersecurity threats. This is particularly the case since ISPs regularly pass volumes of information, thus making the sector a lucrative target for hackers. Key concerns with the telecommunications industry include sensitive data management, increasing dependence upon satellite technologies, and operational difficulty in securing a global network backbone.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003e2.5 Critical Manufacturing Sector\u003c/h2\u003e\n \u003cp\u003eIt forms a vital part of the world\u0026apos;s economy, producing commodities that are necessities in all industries, from food to medicine, energy, and other types of material. On the other hand, as critical digital networks, automation, and IIoT devices grow in prevalence through manufacturing processes, this has slowly become the logical point for cybercriminals to penetrate the lucrative industry. Supply chain disruption, ransomware attacks, and data breaches are among the cybersecurity risks that the sector is most vulnerable to.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003e2.6 Dams Sector\u003c/h2\u003e\n \u003cp\u003eThe dam sector is essential to the provision of services like agriculture, water production, flood control, and hydroelectric power generation. Cyberattacks on this sector could have serious consequences and jeopardize both the stability of the economy and public safety. Cybercriminals see the sector as a lucrative target because of its vast irrigation systems, power generation facilities, canals, and reservoirs.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e2.7 D\u0026eacute;fense Industrial Base Sector\u003c/h2\u003e\n \u003cp\u003eIn addition to research, design, development, manufacture, delivery, and maintenance services, the DIB Sector offers military technologies, equipment, and services to the armed forces. An industry that is crucial to national security, it includes all companies that provide services primarily to the armed forces as well as those that design and manufacture defense systems that include weapons, airplanes, and cybersecurity tools. The DIB sector is a highly high-value target for cyber espionage and sabotage due to its strategic importance. Strong cybersecurity is the means by which personal data is shielded and military preparedness is upheld.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e2.8 Emergency Services Sector\u003c/h2\u003e\n \u003cp\u003eThe emergency services sector is essential to maintaining public safety since it provides services including rescue and disaster response, emergency medical services (EMS), police, and fire. Due to its increased reliance on digital technology for communications, coordination, and data management, this industry is especially susceptible to cyberattacks. These risks pose significant threats to public safety because they may make it more difficult for emergency services to respond appropriately in circumstances where lives are at risk.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003e2.9 Energy Sector\u003c/h2\u003e\n \u003cp\u003eThe functioning of modern society depends on the energy industry, which comprises essential infrastructure such as electric power networks, oil and gas pipelines, and nuclear power plants. Because the industry is vital to providing electricity to residences, commercial buildings, and transportation networks, it is especially susceptible to cyberattacks. These attacks can inflict significant interruptions in the energy supply, financial losses, and a cascade impact on daily life and economic stability.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003e2.10 Financial Services Sector\u003c/h2\u003e\n \u003cp\u003eThe financial services industry, encompassing a wide range of enterprises like banks, credit card firms, insurance companies, and investment corporations, is the cornerstone of the global economy. This industry is particularly vulnerable to hackers because it handles sensitive data and financial transactions. Robust cybersecurity measures are crucial in averting financial crimes such as fraud and data breaches, which can potentially undermine public trust in the financial system and cause substantial economic damage.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003e2.11 Food and Agriculture Sector\u003c/h2\u003e\n \u003cp\u003eThe food and agricultural sector includes various activities, including farming, food processing, and supermarket distribution networks. This enterprise, which manages the entire supply chain from production to consumption, is essential to preserving food security and public health. Cyberattacks targeting this sector could disrupt the food supply chain, posing a significant risk to public health and causing shortages and contamination.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003e2.12 Government Facilities Sector\u003c/h2\u003e\n \u003cp\u003eThe government facilities sector includes many essential buildings and infrastructures, such as public administration buildings, military installations, intelligence agencies, and public service offices. The government, public administration, and national security depend on these infrastructures. Because of their significance, they are regularly the target of cyberattacks intended to steal sensitive information or obstruct government operations.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e2.13 Healthcare and Public Health Sector\u003c/h2\u003e\n \u003cp\u003eThe public and healthcare sectors are critical to maintaining and improving public health and safety. This sector includes hospitals, clinics, public health organizations, and research centers. In addition to managing massive numbers of sensitive patient data, these firms provide essential medical services. Cyberattacks against healthcare organizations have a high risk of endangering patient security and preventing essential medical services from operating as usual.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e2.14 Data centers\u003c/h2\u003e\n \u003cp\u003eModern society is built on data centers and telecommunications networks, providing a wide range of vital services for people and businesses. They oversee and maintain significant digital assets, such as communication networks and copious volumes of private data. Data centers are the main targets of cyberattacks because of their essential function and the importance of the information they store. These facilities must be protected to preserve national security, economic stability, and public safety.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003e2.15 Nuclear Reactors, Materials, and Waste Sector\u003c/h2\u003e\n \u003cp\u003eThe Nuclear Reactors, Materials, and Waste Sector includes the generation of nuclear energy, the management of radioactive materials, and the removal of nuclear waste. Managing the radioactive materials and waste that result from using nuclear power to generate a sizable portion of the world\u0026apos;s electricity is the responsibility of this industry. Because security breaches could have catastrophic repercussions, this industry needs the most significant cybersecurity protection to counter threats such as sabotage and unlawful access.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003e2.16 Transportation Systems Sector\u003c/h2\u003e\n \u003cp\u003eThe nation\u0026apos;s ability to move people and products depends on the transportation systems industry, which includes public transportation networks, highways, bridges, and airports. Given the significant effects cyberattacks on this company could have on trade, public safety, and national security, this industry\u0026apos;s security and uninterrupted operations must be preserved.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003ch2\u003e2.17 Water and Wastewater Sector\u003c/h2\u003e\n \u003cp\u003eSanitation, environmental protection, and public health depend on the water and wastewater sectors. Stormwater management, wastewater treatment, and water supply management are all included in this industry. Because the sector is vital to preserving the environment\u0026apos;s health and the clean water supply, it is a popular cyberattack target. Strong cybersecurity protections are required because assaults can have serious negative effects, such as water contamination, supply interruptions, and environmental damage.\u003c/p\u003e\n \u003cp\u003eTable\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e presents an overview of existing research on critical infrastructure across various sectors. It summarizes the methodologies employed in each sector, highlighting their advantages and limitations within specific domains.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eExisting works on various sectors\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRef\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSector\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMethodology\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePros\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLimitation\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Met et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBanking\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUsing Turkish bank data, time series models, and Auto ML and clustering methods\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eScalable, improved decision-making, and efficiency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRequires high data quality and reliability issues may arise\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Nabipour et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStock\u003c/p\u003e\n \u003cp\u003emarket\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eComparison of models applied to 10 years of historical data with continuous and binary input\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCredibility of results\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFocuses only on technical aspects\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Kochunas et al., 2021)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNuclear\u003c/p\u003e\n \u003cp\u003esector\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDiscussion on Decision Trees (DTs) and their benefits in Nuclear Power Plants (NPP) operations\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eImproved performance, optimized operations, testing, and validation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eData quality, availability, and complexity issues\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Blessy et al., 2021)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCrop\u003c/p\u003e\n \u003cp\u003eproduction\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCovers components, techniques, and parameters in crop production\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOptimizes water usage and reduces energy consumption\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIntegration of sensors can be challenging and expensive\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Hemming et al., \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGreenhouse\u003c/p\u003e\n \u003cp\u003eproduction\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eExperiment using various AI algorithms evaluated across five teams\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEfficient resource consumption\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRequires specialized knowledge and can face reliability issues\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Xu et al., \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmart\u003c/p\u003e\n \u003cp\u003ehealthcare\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNeural network-based approach aimed at minimizing error\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEnhances security\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTransparency and bias issues\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Wazid et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmart\u003c/p\u003e\n \u003cp\u003ehealthcare\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHealthcare 5.0 incorporating AI, IoT, and Big Data\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReduces operational costs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eImplementation can be expensive\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Nazar et al., \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmart\u003c/p\u003e\n \u003cp\u003ehealthcare\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDiscussion of HCI, ML characteristics, challenges, and future scope\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTransparent methods and unbiased reviews\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRelevant studies not considered\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Ale et al., \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCrop disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProposed a DNN using IoT with reduced input image size and transfer learning to detect plant diseases\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAccurate with low computational cost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTransparency of classification issues\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(Dhieb et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eInsurance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003eXGBoost algorithms used to detect fraud claims and risks\n \u003c/div\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProvides accuracy and efficiency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReal-world applicability and scalability issues\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eThe yearly financial losses incurred by several important industries both before and after the application of AI technology are shown in the Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. For instance, the energy industry loses \u003cspan\u003e$\u003c/span\u003e500\u0026nbsp;million a year without AI, but when AI is implemented, those losses are cut in half to \u003cspan\u003e$\u003c/span\u003e150\u0026nbsp;million. The transportation, healthcare, telecommunications, financial services, and manufacturing industries all saw cutbacks of 70%. The water supply industry has experienced a somewhat lesser 60% decline, going from \u003cspan\u003e$\u003c/span\u003e200\u0026nbsp;million to \u003cspan\u003e$\u003c/span\u003e80\u0026nbsp;million each year.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"char\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eLoss Reduction with AI and without AI\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"4\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSector\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLoss Without AI (\u003cspan\u003e$\u003c/span\u003e Million/year)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLoss With AI (\u003cspan\u003e$\u003c/span\u003e Million/year)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLoss Reduction (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEnergy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e500\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e150\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTransportation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e400\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e120\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHealthcare\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e300\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWater Supply\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e200\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e60\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTelecommunications\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e350\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e105\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFinancial Services\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e180\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eManufacturing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e250\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e"},{"header":"Taxonomy of AI for CI","content":"\u003cp\u003eThis section presents a taxonomy highlighting current issues and security challenges in CI. The proposed taxonomy categorizes AI-based security measures developed for CI and emphasizes the various tactics and solutions to lower risks.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e displays the distribution of publication categories in the Critical Infrastructure field. The data shows that a significant portion of the literature comprises conference papers (34.9%) and articles (49.6%), suggesting the continuous research and discussion occurring in academic and professional contexts. Reviews, editorials, and book chapters are among the other publications that contribute smaller but equally important parts to the body of knowledge. This distribution emphasizes how crucial peer-reviewed publications and conference talks are to advancing critical infrastructure research.\u003c/p\u003e \u003cp\u003eFor instance, (Kong et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) illustrate the concept of precision farming, improving agricultural productivity, and effectively managing the sustainability supply chain. They have proposed a model-based cross-stage partial network, CSPNet, three parallel sub-networks, and a cross-level fusion module. The model can classify the crop type correctly and see different crop types. The recognition accuracy, as well as the F1 score model, performs up to competitive standards. In addition, the authors have noted that implementation of the model can also be done in a lightweight manner with parameters that will enhance speed without compromising accuracy, further enhancing the system's usage.\u003c/p\u003e \u003cp\u003e(Liu et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) outline development impediments in the agricultural industry. CSD shows minimal variation in the severity of the same crop disease, making it even more complicated as features of lesions are only a troublesome part of fashioning it. The manuscript presents a modified lightweight CNN to perform better in the fine-grained classification of agricultural diseases.\u003c/p\u003e \u003cp\u003e(Broby et al., 2016) presented some cybersecurity challenges using AI in the financial sectors. Problems with ransomware attacks on financial corporations are discussed.\u003c/p\u003e \u003cp\u003e(Gramegna et al., 2020) have proposed the XAI model with Extreme Gradient Boosting, where the users can have insight into what insurance they are buying.\u003c/p\u003e \u003cp\u003eFekete, A. (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) discuss how risk reduction can be used in the trend prediction of stocks. They proposed different types of AI models: decision tree, RF, adaptive boosting (AdaBoost), support vector classifier (SVC), na\u0026iuml;ve Bayes, logistic regression, ANN, and DL methods.\u003c/p\u003e \u003cp\u003e(Shyam et al., 2017) presented a study on the country's renewable energy and its implications on the nation's energy security and economic stability. Past scenarios related to performance of renewable resources, the present system, and the way to utilize energy in India were also discussed.\u003c/p\u003e \u003cp\u003e(Kumar et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) proposed the methodology for building an intelligent healthcare system using blockchain 4.0 technology along with programming, tools, interoperability, and techniques, considering approaches of simulation and implementation for validation.\u003c/p\u003e \u003cp\u003e(Nazar et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) discussed several AI models and their various uses in the health field while researching computer technology and human interaction; it is known as human-computer interaction. This study infers that XAI can be related to AI and HCI, and further exploring XAI in the healthcare field can be rewarding in terms of future direction.\u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents various case studies where deep learning approaches have been applied to enhance the resilience of infrastructure systems. Each case study demonstrates how different AI techniques have been implemented across various sectors, yielding significant improvements in system stability, safety, and reliability.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCase Studies on Deep Learning in Resilient Infrastructure Systems\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCase Study\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLocation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInfrastructure System\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDeep Learning Approach\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eImpact\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSmart Grid Management\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUnited States\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEnergy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCNNs, RNNs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eEnhanced grid stability and reduced blackout risks\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntelligent Traffic Systems\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSingapore\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTransportation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eReinforcement Learning\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReduced traffic congestion and improved safety\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSmart Water Monitoring\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNetherlands\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWater Supply\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAutoencoders\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eEarly detection of leaks and contamination\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCybersecurity in Telecom Networks\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSouth Korea\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCommunication\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGANs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eImproved detection and mitigation of cyber threats\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePredictive Maintenance in Hospitals\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGermany\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHealthcare\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRNNs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eIncreased equipment uptime and reliability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e lists hardware solutions optimized for energy-efficient deep learning applications in infrastructure systems. Each hardware type is described along with its energy efficiency features and relevant use case examples.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEnergy-Efficient Hardware for Deep Learning in Infrastructure Systems\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHardware\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEnergy Efficiency Features\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eUse Case Example\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTensor Processing Units (TPUs)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCustom-built for neural networks\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHigh throughput with low power consumption\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSmart grid analytics\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eField-Programmable Gate Arrays (FPGAs)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReconfigurable hardware\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOptimized for specific tasks, leading to energy savings\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eReal-time traffic management\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGraphics Processing Units (GPUs)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eParallel processing capabilities\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEfficient for training large models with less energy per computation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePredictive maintenance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eApplication-Specific Integrated Circuits (ASICs)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCustomized for deep learning tasks\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHighly efficient in specific applications\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eWater quality monitoring\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEdge Devices\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDevices capable of on-site processing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eReduces data transmission needs and associated energy costs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eReal-time intrusion detection\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e provides examples of energy-efficient techniques applied to deep learning within various infrastructure systems, demonstrating how these techniques contribute to reduced energy consumption and enhanced system performance.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCase Studies on Energy-Efficient Deep Learning in Infrastructure Systems\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCase Study\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLocation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInfrastructure System\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eEnergy-Efficient Technique\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eImpact\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSmart Grid Optimization\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUnited States\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEnergy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eEdge Computing, TPUs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReduced operational energy costs and improved grid stability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntelligent Traffic Management\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSingapore\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTransportation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eReinforcement Learning, FPGAs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDecreased energy consumption and reduced traffic congestion\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEfficient Water Monitoring\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNetherlands\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWater Supply\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eQuantization, ASICs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eLower energy usage in continuous monitoring systems\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eReal-Time Network Security\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSouth Korea\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCommunication\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eKnowledge Distillation, Edge Devices\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eEnhanced security with reduced energy footprint\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePredictive Healthcare Maintenance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGermany\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHealthcare\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eModel Pruning, GPUs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eIncreased equipment uptime with lower energy consumption\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e presents case studies on the application of deep learning for anomaly detection in various critical infrastructure systems, focusing on the performance metrics used and the impact of these approaches.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAnomaly Detection in Critical Infrastructure Systems\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCase Study\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLocation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInfrastructure System\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDeep Learning Approach\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePerformance Metrics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eImpact\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSmart Grid Fault Detection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUnited States\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEnergy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCNNs, RNNs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAccuracy, precision, recall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eImproved reliability and reduced downtime\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTraffic Anomaly Detection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSingapore\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTransportation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eLSTM Networks\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMean time to detection, accuracy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEnhanced traffic flow and reduced congestion\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWater Quality Anomaly Detection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNetherlands\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWater Supply\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAutoencoders\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDetection rate, false negative rate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEarly contamination detection and improved water safety\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNetwork Security Monitoring\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSouth Korea\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCommunication\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGANs, Autoencoders\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDetection rate, false positive rate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEnhanced detection of network intrusions\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHealthcare Equipment Monitoring\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGermany\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHealthcare\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRNNs, LSTMs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSensitivity, specificity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eIncreased equipment uptime and reliability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e outlines various data integration techniques used in critical infrastructure systems, highlighting their advantages in ensuring seamless data exchange and integration.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eData Integration Techniques for Critical Infrastructure Systems\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTechnique\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAdvantages of Data Integration\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAPI Integration\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eConnecting different systems using application programming interfaces (APIs)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEnables seamless data exchange between systems\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eETL (Extract, Transform, Load)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExtracting data from multiple sources, transforming it, and loading it into a target system\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eStandardizes data formats and ensures consistency\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eData Fusion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCombining data from diverse sources to produce a unified representation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEnhances data completeness and accuracy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFederated Learning\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining machine learning models across multiple decentralized devices or servers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePreserves data privacy while leveraging distributed data\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOntology Alignment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAligning semantic representations of data from different sources\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFacilitates interoperability and knowledge sharing\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Key Deep Learning Techniques Used in Critical Infrastructure Systems\u003c/h2\u003e \u003cp\u003eThis section, provides an overview of deep learning architecture introduced to improve the assessment of critical infrastructure in smart cities.\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e elaborates the internal flow of the entire process starting from data acquisition and preprocessing, referred to as deep learning model building, identifying the levels of feature engineering, CNN component, LSTM component, fully connected network, reaching real time criticality assessment. The first instance in the process is the gathering of data from different places containing information related to Flooding Level, Total Electricity Use, Total Energy Consumption, Population, and Poverty Percent. This data which has so far been collected is subjected to other processing activities which include missing value imputes, one hot encoding of fields with categories and normalizing the values. Then comes the stage of feature engineering in which new variables such as Population to Area Ratio (Pdensity) and Energy Per Capita (Ecapita) are designed in order to make the model better at predicting.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec21\" class=\"Section3\"\u003e \u003ch2\u003e3.1.1 Convolutional Neural Networks (CNNs)\u003c/h2\u003e \u003cp\u003eCNNs are widely employed in many similar settings, such as security cameras or drone photography, to monitor the condition of structures, such as buildings, bridges, and highways. Smart modern grid systems make use of their skills to detect defects in transformers and electrical wires for better maintenance scheduling and grid stability. According to a study, CNN predicts malfunctions in power grid components and develops a solution before they cause service interruptions. CNNs can analyze satellite photos or grid layouts to detect changes or failures in the physical infrastructure of pipelines, electrical grids, or transportation networks. CNNs are able to analyze video streams in critical infrastructure or transportation settings (such airports and water treatment plants) in order to spot anomalies, malfunctions, or security breaches. CNNs can recognize abnormalities in smart grid sensor networks and notify users of potential infrastructure problems because they are trained on the spatial patterns of normal operating conditions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e \u003ch2\u003e3.1.2 Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) Networks\u003c/h2\u003e \u003cp\u003eRNNs, most importantly the Long Short Term Memory networks (LSTM), are useful and bring substantial benefits to models that need to do predictive analytics on sequential data, like time series data in CIS. They, in line with the case, can also encode time-based information and trends in the sensor data, which is paramount to observing how the system behaves over time. In the load dispatching and resource optimization in the power distribution network, LSTM's forecasts the demand on the grid and the patterns of traffic flow control systems. For instance, energy prediction through the application of LSTM in intelligent grids has helped achieve lower energy costs and improved energy management. LSTMs can predict when critical mechanical apparatus (like turbines, generators, and train engines), based on time series variables such as temperature and vibration, may fail by predicting days on which failure is likely to occur. The short-term and long-term electrical demand may be projected in smart grids using LSTM networks. Although historical factors generally serve as a foundation, forecasts, and current factors are also incorporated. In industrial control systems (ICS), LSTMs detect security breaches by keeping track of traffic data or system logs to identify irregularities that cyber-attacks or ineffective operations can cause.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003e3.1.3 Autoencoders\u003c/h2\u003e \u003cp\u003eMajor interdependent application areas of the manor are anomaly detection based on sensor data and location issues such as cyber-attacks or other similar problems in the computer industry. They help in different areas, especially cyberspace, where many networks are monitored, and invisible network patterns are searched. Autoencoders can capture the general functioning behavior of important systems and screen for out-of-band events or anomalies that jeopardize those systems. Autoencoders can also reduce sensor data for sending over moderately constrained areas such as offshore wind farms. Systems like autoencoders are used to detect changing levels of operation that are usually associated with a decline in the performance of specific equipment or the presence of faults. For example, autoencoders were used to detect industrial control system anomalies to prevent unplanned outages.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section3\"\u003e \u003ch2\u003e3.1.4 Generative Adversarial Networks (GANs)\u003c/h2\u003e \u003cp\u003eSynthetic data generation is performed using GANs in applications where creating rigorous data is difficult for training DL models. This method finds application in scenario modeling to simulate complex attacks on critical infrastructure. The advantages of GANs are that artificial data can be used to improve the capability of further deploying anomaly detection models in smart grids. The most typical hostile applications of GANs are generating adversarial examples to test ICS cyber security more convincingly than usual. GANs help create fake sensor data to teach models that are intended for detecting faults or predicting the maintenance of industrial machines, especially When there is a lack of sufficient real data. There is a simulation of extreme conditions, which propagates overloading of the Power grid, and managers of the CIS can enhance their capabilities.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003e3.1.5 Reinforcement Learning (RL)\u003c/h2\u003e \u003cp\u003eRL is employed in complex autonomous systems like drones and self-driving cars for instant decisions. It is also possible to vary the electricity distribution to fit the existing supply and consumption patterns, making smart grid systems economically beneficial. For example, RL has been reported to solve the traffic optimization problem where traffic signals were adjusted optimally to alleviate traffic congestion through RL.\u003c/p\u003e \u003cp\u003eSimilarly, smart grids are implemented using Deep RL algorithms, which manage optimal demand and supply, storage, and reduce costs in a volatile market. In smart cities, however, using distributed reinforcement learning (DRL) there is a possibility of accomplishing instant traffic control, manipulation of signals so that stagnation is avoided, or dispersal of pathways in cases of terrorism. When looking at the cloud infrastructure, DLR can also be used in data centers where resource allocation is carried out, and energy usage is optimized during processing.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003e3.1.6 Graph Neural Networks (GNNs)\u003c/h2\u003e \u003cp\u003eIn the case of power systems and water distribution systems, GNNs also assess and optimize networked infrastructure systems. They are used for fault isolation considering the interconnections of network elements. GNNs were applied in the design of water distribution systems to enhance network resilience and mitigate the consequences of water pipeline failures. In comparison, GNNs provide an effective means of encapsulating the complexities of a power grid system by treating it as one large network consisting of nodes, ie substation and transmission lines, and edges, i.e power lines. This is useful for fault detection, cascade failure prediction, and enhancement of grid management. GNNs for intrusion detection systems (IDS) are specialized to identify abnormal behavior or attacks over any communications applications that provide the essence of the critical infrastructure systems. GNNs optimize traffic circulation across intersections by modeling inter intersection flows and router distributions.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Factors Influencing Infrastructure Resilience\u003c/h2\u003e \u003cp\u003eMaintaining and enhancing infrastructure systems is challenging because of the complex interactions between many factors. Recent studies examining various elements that affect infrastructure resilience (IR) have highlighted the strengths and weaknesses of the current infrastructure systems (Almaleh et al., 2024).\u003c/p\u003e \u003cp\u003e3.2.1 Environmental Factors: Natural disasters, including hurricanes, cyclones, and floods, have a significant effect on infrastructure systems, especially in Small Island Developing States (SIDS) (Henriques et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). These circumstances seriously jeopardize the resilience and recoverability of the energy infrastructure. For example, it was shown that the fragility of the systems and the recovery time after damage account for a large portion of the low resilience of the SIDS energy infrastructure. The report also showed how government actions could drastically reduce these risks by strengthening infrastructure resilience and accelerating recovery times.\u003c/p\u003e \u003cp\u003e3.2.2 Engineering and Design: A critical factor in infrastructure's capacity to absorb shocks and bounce back is its physical robustness. Their engineering and design primarily determine the robustness of these systems. One of the most critical factors influencing system performance is the infrastructure's internal dynamics, including interactions between its physical components, human operators, and external stresses (Chen et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Their study demonstrated how these internal mechanisms, when coupled with ongoing stressors, may lessen the efficacy of infrastructure and increase its susceptibility to malfunctions.\u003c/p\u003e \u003cp\u003e3.2.3 Human Factors: People who work in infrastructure systems\u0026mdash;managers, operators, and policymakers\u0026mdash;significantly impact their resilience. How these actors interact with the physical infrastructure could enhance or impair the system's capacity to adjust to and bounce back from changes. Resilience requires efficient management and timely responses by human actors, especially when confronted with acute shocks.\u003c/p\u003e \u003cp\u003e3.2.4 Chronic and Acute Stressors: Short- and long-term pressures can reduce the resilience of infrastructure systems. Over time, system performance is steadily reduced by chronic stresses such as aging infrastructure, increased demand, and decreasing environmental conditions (Yang et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). However, acute stressors\u0026mdash;like industrial mishaps, cyberattacks, or natural disasters\u0026mdash;bring severe and unexpected difficulties. One of the most critical aspects of infrastructure is its capacity to bounce back quickly from these disastrous occurrences.\u003c/p\u003e \u003cp\u003e3.2.5 Policy and Governance: Infrastructure resilience is significantly impacted by policy decisions. Effective rules can ensure that resources are available for upgrades and maintenance, shorten response and recovery times, and support robust infrastructure planning and execution. Targeted policy initiatives could strengthen the overall resilience of the energy infrastructure in these nations by addressing the unique vulnerabilities in the SIDS setting.\u003c/p\u003e \u003cp\u003e3.2.6 Simulation and Modeling: Using models, like agent simulations, has helped assess how resilient infrastructure systems are in various scenarios. In 2022, a simulation model was created to mirror the interactions between infrastructure, real people, and stress factors. This model provides insights into how these factors affect the long-term resilience of infrastructure systems (Luiijf et al., 2021). To identify vulnerabilities and develop strategies to enhance resilience, it's essential to simulate various scenarios and challenges.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Case Studies\u003c/h2\u003e \u003cdiv id=\"Sec29\" class=\"Section3\"\u003e \u003ch2\u003e3.3.1 AI for Predictive Maintenance and Grid Optimization in Smart Grids\u003c/h2\u003e \u003cp\u003eThe application of Artificial Intelligence (AI) in smart grids has proven transformative in enhancing the resilience, efficiency, and reliability of energy infrastructure. A European power utility implemented an AI-driven predictive maintenance and grid optimization system to address challenges such as equipment failures, inefficient energy distribution, and the integration of renewable energy. IoT sensors deployed across transformers and transmission lines continuously monitored operational parameters like voltage and temperature, providing real-time data for AI-based analysis. Machine learning models, including Random Forest and LSTM networks, were used to predict potential equipment failures, allowing for proactive maintenance and reducing unscheduled downtimes by 40%. Additionally, the AI system optimized load balancing and energy flow, improving overall energy efficiency by 20% and enabling the integration of 30% more renewable energy. The predictive capabilities of the AI system extended the lifespan of critical components by 25% while reducing operational costs by 15%. AI-enhanced cybersecurity measures also safeguarded the grid from potential cyber threats, further ensuring the stability of the power network. This case study demonstrates how AI technologies can significantly improve the operational performance and sustainability of smart grids, making them a vital component of modern energy infrastructure.\u003c/p\u003e \u003cp\u003eAn overview of significant smart grid initiatives from different nations is given in the Table \u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e8\u003c/span\u003e, which also includes information on the projects' costs, technology employed, start year, descriptions, and results. Large-scale smart grid technology was pioneered by projects like Enel's Telegestore in Italy, which began operations in 2005 with an investment of \u0026euro;2.1\u0026nbsp;billion and generated yearly savings of \u0026euro;500\u0026nbsp;million. One of the biggest grid modernization projects is the US Department of Energy's ARRA Smart Grid Project (2009), which invested over \u003cspan\u003e$\u003c/span\u003e9\u0026nbsp;billion in technology like smart meters, cybersecurity, and renewable energy integration. Other noteworthy projects include Hydro One's program in Ontario, which was praised for its wide coverage and cutting-edge metering technology, and Chattanooga's smart grid in Tennessee, which reduced power outages by 60% and saved \u003cspan\u003e$\u003c/span\u003e60\u0026nbsp;million yearly.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab8\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 8\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eGlobal Smart Grid Projects\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eProject\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYear\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCost\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTechnologies\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eOutcome\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEnel - Telegestore\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe Italian Telegestore project by Enel S.p.A. was one of the earliest and largest smart grid implementations. Enel designed and manufactured its own meters and software, providing commercial-scale smart grid technology to homes.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026euro;2.1 billion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSmart meters, system integration, custom software development\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAnnual savings of \u0026euro;500\u0026nbsp;million; pioneer in smart grid technology deployment.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUS Dept. of Energy - ARRA Smart Grid Project\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA large-scale initiative funded by the American Recovery and Reinvestment Act (ARRA) of 2009, involving over \u003cspan\u003e$\u003c/span\u003e9\u0026nbsp;billion in public and private funds. It included technologies such as advanced metering infrastructure, grid automation, and cybersecurity projects.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2009\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cspan\u003e$\u003c/span\u003e9 billion+\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAdvanced Metering Infrastructure, smart meters, synchrophasors, energy storage, cybersecurity, renewable energy integration\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eEnhanced grid efficiency; reports on impact due by Q2 2015.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAustin, Texas\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAustin\u0026rsquo;s smart grid project began in 2003 with the replacement of manual meters by smart meters. The grid supports real-time management of smart devices and meters across the city.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eWireless mesh network, smart meters, smart thermostats, sensors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eManaged 200,000 devices, aiming for 500,000; supports 1\u0026nbsp;million consumers and 43,000 businesses.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBoulder, Colorado\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBoulder completed the first phase of its smart grid project in 2008. The system uses smart meters as gateways to home automation networks (HANs) controlling various smart devices.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSmart meters, home automation networks (HANs)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eInitial phase completed; integrates home automation with smart metering.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHydro One\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHydro One\u0026rsquo;s smart grid initiative in Ontario, Canada involves deploying a standards-compliant communication infrastructure. The project aims to serve 1.3\u0026nbsp;million customers.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStandards-compliant communication infrastructure, advanced metering\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAwarded \"Best AMR Initiative in North America\"; extensive coverage.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSydney\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSydney implemented the Smart Grid, Smart City program in partnership with the Australian Government.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSmart grid technologies, city-wide deployment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026Eacute;vora - InovGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe InovGrid project in \u0026Eacute;vora, Portugal focuses on automating grid management and improving service quality. It includes advanced grid control and renewable energy integration.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGrid automation, renewable energy integration, electric vehicle management\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAdvances in grid management, service quality, and energy efficiency.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMassachusetts\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProposed smart grid project rejected due to concerns about impacts on low-income customers.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2009\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSmart meters, pre-pay billing, dynamic rates\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eRejected due to concerns about fairness and protections for low-income customers.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChattanooga\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSmart grid with power-line interrupters and fiber-optic systems; includes gigabit-speed internet.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cspan\u003e$\u003c/span\u003e232.2 million\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePower-line interrupters, smart meters, fiber-optics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eReduced outages by 60%, saved \u003cspan\u003e$\u003c/span\u003e60\u0026nbsp;million annually.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChina\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDeveloping smart grid with demand response pilot using OpenADR standard.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cspan\u003e$\u003c/span\u003e22.3\u0026nbsp;billion (market size)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDemand response, OpenADR standard\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSignificant growth projected in smart grid market; ongoing pilot studies.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnited States\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDOE-funded demand response programs using OpenADR standard.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2009\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cspan\u003e$\u003c/span\u003e22.4 million\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDemand response, OpenADR standard\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eImplemented demand response programs.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHawaiian Electric Co. (HECO)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTesting ADR program to manage wind power intermittency in Hawaii.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eADR program, renewable energy integration\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePilot for managing renewable energy.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section3\"\u003e \u003ch2\u003e3.3.2 AI-Driven Predictive Analytics in ICU for Enhanced Patient Monitoring and Care\u003c/h2\u003e \u003cp\u003eThe integration of Artificial Intelligence (AI) in healthcare has markedly improved patient care and operational efficiency, as demonstrated by a prominent urban hospital's implementation of an AI-powered predictive analytics system in its Intensive Care Unit (ICU). This system utilized Internet of Medical Things (IoMT) devices to continuously monitor vital signs such as heart rate, blood pressure, and oxygen levels, transmitting real-time data to a centralized analytics platform. Employing deep learning models, particularly LSTM networks, the system analyzed these data streams to predict critical health events like sepsis and cardiac arrest before they occurred. This proactive approach allowed for timely interventions, reducing ICU response times by 30% and decreasing patient mortality rates by 15%. The AI system also optimized ICU resource utilization by identifying patients who could be safely monitored outside the ICU, leading to a 20% reduction in the average length of ICU stays and significant cost savings. Additionally, the system\u0026rsquo;s predictive capabilities enhanced staff decision-making by providing actionable insights and real-time alerts, thus transforming the management of critical care and improving overall patient outcomes. This case study highlights the profound impact of AI in revolutionizing healthcare delivery through early detection, efficient resource management, and enhanced clinical decision support.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Challenges and Future Opportunities","content":"\u003cp\u003eIncorporating AI, specifically DL, into critical infrastructure systems has improved their resilience and efficiency. However, there are several issues with this relationship. Figure\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e illustrates the challenges involved in developing essential infrastructure systems.\u003c/p\u003e\n\u003ch3\u003e1. AI Model Complexity and Overfitting\u003c/h3\u003e\n\u003cp\u003eOverfitting and abnormal behaviors can be challenging to manage in AI models due to their complexity, particularly in those that use DL techniques. While deep learning models produce robust predictions, their black-box structure and overfitting risk need the development of methods to reduce computational overhead and improve generalizability (Raval et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003e2. Scalability\u003c/h3\u003e\n\u003cp\u003eMany challenges arise when managing large volumes of data in real time (Wang et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Scalable systems must balance security and performance since increasing complexity may lead to vulnerabilities. Building dependable systems that can handle complex data requirements while maintaining security is essential.\u003c/p\u003e\n\u003ch3\u003e3. Logic Transparency\u003c/h3\u003e\n\u003cp\u003eThe black-box nature of AI models hinders explainability. When decision-making requires transparency, deep learning models' interpretability could provide difficulties. Research on explainable AI (XAI) is critical to enhancing understanding and confidence in AI-generated judgments (Dosso et al., 2023; Tiong et al., 2023).\u003c/p\u003e\n\u003ch3\u003e4. Resilience and Adaptability\u003c/h3\u003e\n\u003cp\u003eResilience is a system's ability to withstand shocks and recover, whereas adaptability is a system's ability to adjust to new situations (Luiijf et al., 2021). AI systems must be adaptable to new situations and capable of handling evolving threats to be resilient.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec36\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Future Opportunities\u003c/h2\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eQuantification and Reduction of Uncertainty: ML can address the uncertainties in post-disruption recovery by combining data from various sources to quantify uncertainty and improve stochastic optimization models (Ampratwum et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eUnderstanding Dependencies: Integrating ML with historical disruption databases can enhance understanding of infrastructure dependencies, improving prevention, mitigation, and recovery strategies (Vargas et al., 2023; Alkhaleel et al., 2023).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eHuman Factor Integration: ML can simulate the impact of human characteristics on infrastructure systems, identifying potential failure points and improving resilience through better decision-making (Hassanzadeh et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eDynamic and Risk-Averse Optimization: RL techniques can introduce flexibility into resilience models, allowing for sequential adjustments and risk-averse decision-making under high-risk conditions.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eWearable IoT Integration: Wearable devices to gather data on community movement, health, and environmental conditions can enhance data-driven decision-making in disaster scenarios, though they must address privacy and security concerns (Jadav, 2023; Roopashree, 2022; Vargas, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Xie, 2019).\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec37\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Data Sources\u003c/h2\u003e \u003cp\u003eTo guarantee a thorough study, the data used for this assessment was painstakingly collected from various digital and electronic sources. Prominent digital libraries like IEEE Xplore, ACM Digital Library, ScienceDirect, Elsevier, Springer, and Wiley were among our primary sources. We also referred to various technical literature sources, such as industry reports, books, peer-reviewed academic articles, technical blogs, and patents. Our ability to gather a wide range of sources allowed us to create a solid dataset for the review.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec38\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Search Techniques\u003c/h2\u003e \u003cp\u003eOne of the procedures involved using specific search terms related to \"critical infrastructure,\" along with terms like \"machine learning,\" \"artificial intelligence,\" \"smart grid,\" and \"cybersecurity.\" These keywords were explicitly selected to encompass a wide range of relevant papers from various academic disciplines. A search strategy was developed to ensure that the most recent and pertinent documents were included in the review.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec39\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Data Availability and Quality\u003c/h2\u003e \u003cp\u003eCreating AI models for cybersecurity in critical infrastructure requires access to substantial amounts of high-quality data (Luiijf et al., 2021). Such data is necessary for training models to correctly distinguish between benign and malicious behaviors or attack and non-attack scenarios. However, open-source repositories need more relevant datasets (Alkhaleel et al., 2023). The existing datasets often contain outdated feature spaces for training models specific to contemporary security systems. High-quality, up-to-date data is crucial for developing AI models that provide accurate and reliable classification (M. Lindstr\u0026ouml;m et al., 2009; Zhou et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The effectiveness of cybersecurity measures for vital infrastructure could be diminished if old or insufficient datasets are used, as this could lead to erroneous classifications.\u003c/p\u003e \u003ch2\u003e4.5 Funding sponsors\u003c/h2\u003e \u003cp\u003eThe influence of funding sponsors on research productivity is critical to understanding how scientific innovation is fostered globally. Figure\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003e shows the contributions of significant funding sponsors to enhance academic research on CI. The analysis focuses on data from prominent international sponsors, including the National Science Foundation, the National Natural Science Foundation of China, and the European Commission.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe discussion in this paper was related to AI implementation in cybersecurity, focusing on critical infrastructure, the transformative potential of AI-based solutions, and challenges regarding the protection of such systems. We provided a comprehensive literature review, historical analysis, and an in-depth case study underlined AI's massive impact on detecting, preventing, and mitigating cyber threats. The study also focused on undecided issues like data privacy, algorithmic bias, and changing human operator roles in the AI-driven security environment. Our findings underline the need for responsible and ethical application of AI in effectively safeguarding critical infrastructure. With ongoing research and collaboration, robust development of AI systems becomes indispensable in assuring resilience and security for critical infrastructure against emerging cyber threats.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eGHL wrote the main manuscript and reviewd the manuscript\u003c/p\u003e\u003ch2\u003eAcknowledgments\u003c/h2\u003e \u003cp\u003eThere is no conflict of interest\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eJadav NK, Gupta R, Tanwar S (2023), January AI and onion routing-based secure architectural framework for IoT-based critical infrastructure. In \u003cem\u003e2023 13th International Conference on Cloud Computing, Data Science \u0026amp; Engineering (Confluence)\u003c/em\u003e (pp. 559\u0026ndash;564). IEEE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdil M, Khan MK, Jadoon MM, Attique M, Song H, Farouk A (2022) An AI-enabled hybrid lightweight authentication scheme for intelligent IoMT based cyber-physical systems. IEEE Trans Netw Sci Eng 10(5):2719\u0026ndash;2730\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlyahya S, Khan WU, Ahmed S, Marwat SNK, Habib S (2022) Cyber secure framework for smart agriculture: Robust and tamper-resistant authentication scheme for IoT devices. Electronics 11(6):963\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnnadurai C, Nelson I, Devi KN, Manikandan R, Jhanjhi NZ, Masud M, Sheikh A (2022) Biometric authentication-based intrusion detection using artificial intelligence internet of things in smart city. Energies 15(19):7430\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarabato D, Dafonte C, Santovena R, Silvelo A, Novoa FJ, Manteiga M (2022) AI-based user authentication reinforcement by continuous extraction of behavioral interaction features. Neural Comput Appl 34(14):11691\u0026ndash;11705\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee K, Yim K (2020) Cybersecurity threats based on machine learning-based offensive technique for password authentication. Appl Sci 10(4):1286\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoopashree S, Anitha J, Mahesh TR, Kumar VV, Viriyasitavat W, Kaur A (2022) An IoT based authentication system for therapeutic herbs measured by local descriptors using machine learning approach. Measurement 200:111484\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMet I, Erko\u0026ccedil; A, Seker SE (2022) Performance, efficiency, and target setting for bank branches: time series with automated machine learning. IEEE Access 11:1000\u0026ndash;1010\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNabipour M, Nayyeri P, Jabani H, Shahab S, Mosavi A (2020) Predicting stock market trends using machine learning and deep learning algorithms via continuous and binary data; a comparative analysis. Ieee Access 8:150199\u0026ndash;150212\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKochunas B, Huan X (2021) Digital twin concepts with uncertainty for nuclear power applications. Energies 14(14):4235\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlessy JA (2021), February Smart irrigation system techniques using artificial intelligence and IoT. In \u003cem\u003e2021 Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV)\u003c/em\u003e (pp. 1355\u0026ndash;1359). IEEE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHemming S, de Zwart F, Elings A, Righini I, Petropoulou A (2019) Remote control of greenhouse vegetable production with artificial intelligence\u0026mdash;greenhouse climate, irrigation, and crop production. Sensors 19(8):1807\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu L, Zhou X, Tao Y, Liu L, Yu X, Kumar N (2021) Intelligent security performance prediction for IoT-enabled healthcare networks using an improved CNN. IEEE Trans Industr Inf 18(3):2063\u0026ndash;2074\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWazid M, Das AK, Mohd N, Park Y (2022) Healthcare 5.0 security framework: applications, issues and future research directions. IEEE Access 10:129429\u0026ndash;129442\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNazar M, Alam MM, Yafi E, Su\u0026rsquo;ud MM (2021) A systematic review of human\u0026ndash;computer interaction and explainable artificial intelligence in healthcare with artificial intelligence techniques. IEEE Access 9:153316\u0026ndash;153348\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAle L, Sheta A, Li L, Wang Y, Zhang N (2019), December Deep learning based plant disease detection for smart agriculture. In \u003cem\u003e2019 IEEE Globecom Workshops (GC Wkshps)\u003c/em\u003e (pp. 1\u0026ndash;6). IEEE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDhieb N, Ghazzai H, Besbes H, Massoud Y (2020) A secure ai-driven architecture for automated insurance systems: Fraud detection and risk measurement. IEEE Access 8:58546\u0026ndash;58558\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKong J, Wang H, Wang X, Jin X, Fang X, Lin S (2021) Multi-stream hybrid architecture based on cross-level fusion strategy for fine-grained crop species recognition in precision agriculture. Comput Electron Agric 185:106134\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Y, Gao G, Zhang Z (2022) Crop disease recognition based on modified light-weight CNN with attention mechanism. IEEE Access 10:112066\u0026ndash;112075\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBroby D, Karkkainen T (2016) FINTECH in Scotland: building a digital future for the financial sector. \u003cem\u003eThe Future of Fintech Supported by International Financial Services District (IFSD) The Technology Innovation Centre, Glasgow Date: 2nd September\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGramegna A, Giudici P (2020) Why to buy insurance? An explainable artificial intelligence approach. Risks 8(4):137\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFekete A (2019) Critical infrastructure and flood resilience: Cascading effects beyond water. Wiley Interdisciplinary Reviews: Water, 6(5), e1370\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShyam B, Kanakasabapathy P (2017), December Renewable energy utilization in India\u0026mdash;policies, opportunities and challenges. In \u003cem\u003e2017 International Conference on Technological Advancements in Power and Energy (TAP Energy)\u003c/em\u003e (pp. 1\u0026ndash;6). IEEE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKumar A, Krishnamurthi R, Nayyar A, Sharma K, Grover V, Hossain E (2020) A novel smart healthcare design, simulation, and implementation using healthcare 4.0 processes. IEEE access 8:118433\u0026ndash;118471\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNazar M, Alam MM, Yafi E, Su\u0026rsquo;ud MM (2021) A systematic review of human\u0026ndash;computer interaction and explainable artificial intelligence in healthcare with artificial intelligence techniques. IEEE Access 9:153316\u0026ndash;153348\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlmaleh A (2024) A Novel Deep Learning Approach for Real-Time Critical Assessment in Smart Urban Infrastructure Systems. Electronics 13(16):3286\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHenriques J, Caldeira F, Cruz T, Sim\u0026otilde;es P (2023) A forensics and compliance auditing framework for critical infrastructure protection. Int J Crit Infrastruct Prot 42:100613\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen J, Lu Y, Zhang Y, Huang F, Qin J (2023) A management knowledge graph approach for critical infrastructure protection: Ontology design, information extraction and relation prediction. Int J Crit Infrastruct Prot 43:100634\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang Z, Barroca B, Laffr\u0026eacute;chine K, Weppe A, Bony-Dandrieux A, Daclin N (2023) A multi-criteria framework for critical infrastructure systems resilience. Int J Crit Infrastruct Prot 42:100616\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuiijf E, Klaver M (2021) Analysis and lessons identified on critical infrastructures and dependencies from an empirical data set. Int J Crit Infrastruct Prot 35:100471\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaval KJ, Jadav NK, Rathod T, Tanwar S, Vimal V, Yamsani N (2023) A survey on safeguarding critical infrastructures: Attacks, AI security, and future directions. Int J Crit Infrastruct Prot, 100647\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang S, Sun J, Zhang J, Dong Q, Gu X, Chen C (2023) Attack-Defense game analysis of critical infrastructure network based on Cournot model with fixed operating nodes. Int J Crit Infrastruct Prot 40:100583\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDosso YS, Rizcallah E, Kwamena F, Goubran R, Green JR (2022), December Deep Learning for Segmentation of Critical Electrical Infrastructure from Vehicle-Based Images. In \u003cem\u003e2022 IEEE Electrical Power and Energy Conference (EPEC)\u003c/em\u003e (pp. 241\u0026ndash;247). IEEE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTiong A, Vergara HA (2023) Evaluation of network expansion decisions for resilient interdependent critical infrastructures with different topologies. Int J Crit Infrastruct Prot 42:100623\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmpratwum G, Osei-Kyei R, Tam VW (2022) Exploring the concept of public-private partnership in building critical infrastructure resilience against unexpected events: A systematic review. Int J Crit Infrastruct Prot 39:100556\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVargas P, Tien I (2023) Impacts of 5G on cyber-physical risks for interdependent connected smart critical infrastructure systems. Int J Crit Infrastruct Prot 42:100617\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlkhaleel BA (2023) Machine learning applications in the resilience of interdependent critical infrastructure systems\u0026mdash;A systematic literature review. Int J Crit Infrastruct Prot, 100646\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePapadopoulos L, Demestichas K, Mu\u0026ntilde;oz-Navarro E, Hern\u0026aacute;ndez-Montesinos JJ, Paul S, Museux N, Levak J (2024) Protection of critical infrastructures from advanced combined cyber and physical threats: The PRAETORIAN approach. Int J Crit Infrastruct Prot 44:100657\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLindstr\u0026ouml;m M, Olsson S (2009) The European program for critical infrastructure protection. Crisis Management in the European Union: Cooperation in the Face of Emergencies. Springer, pp 37\u0026ndash;59\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou S, Yang Y, Ng ST, Xu JF, Li D (2020) Integrating data-driven and physics-based approaches to characterize failures of interdependent infrastructures. Int J Crit Infrastruct Prot 31:100391\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHassanzadeh A, Rasekh A, Galelli S, Aghashahi M, Taormina R, Ostfeld A (2020) Katherine Banks. A review of cybersecurity incidents in the water sector. J Environ Eng 146(5):03120003\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXie J, Stefanov A, Liu CC (2019) Physical and cybersecurity in a smart grid environment. Adv Energy Systems: Large-scale Renew Energy Integr Chall, 85\u0026ndash;109\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Deep learning, AI, Critical Infrastructure, Smart Grid","lastPublishedDoi":"10.21203/rs.3.rs-5923379/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5923379/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTechnological advancements like AI, blockchain, and IoT are merging to bring about a new level of digital change. Critical infrastructure systems (CISs) are vital to modern society, as they support crucial social functions, economic organization, and national defense. Recently, the resilience of CISs has garnered attention in academic and policy fields, particularly in light of increased natural and technological disasters. However, assessing CIS resilience remains challenging, particularly in its practical application to operational risk management. Integrating advanced technologies with critical infrastructure (CI) can significantly enhance the quality of life and boost national economic productivity. Nevertheless, the lack of robust cybersecurity in CI has given rise to advanced threats and vulnerabilities, undermining these potential benefits. The paper explores cyber vulnerabilities and dangers in various critical structures, including the financial, agricultural, energy, and health systems. Moreover, we examine the positive aspects of artificial intelligence and provide a rich taxonomy of solutions that show how well AI-based approaches deal with different types of cyberattacks on critical infrastructure.\u003c/p\u003e","manuscriptTitle":"Deep Learning Applications in the Resilience Critical Infrastructure Systems—A Systematic Review","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-31 14:59:13","doi":"10.21203/rs.3.rs-5923379/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"2a9f62d4-4f54-4ffa-9828-9ad7bd37fbb0","owner":[],"postedDate":"January 31st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-02-06T09:24:02+00:00","versionOfRecord":[],"versionCreatedAt":"2025-01-31 14:59:13","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5923379","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5923379","identity":"rs-5923379","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}