Cyber Threat Using Deep Learning

Cyber Threat Using Deep Learning | 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 Cyber Threat Using Deep Learning Ch. Meghana, B. Vaishnavi, Shaik. Ayazul Hasan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6625113/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 The network faces dangerous incidents from both inside and outside sources that destroy computer systems. The array defense system features intrusion detection as one of its active elements which protects network security from undetected problems. Traditional intrusion systems face difficulties with detection precision and rate along with generating several false alerts that require extra resources to identify enlightened attack patterns. An ensemble-based breach detection serves as the proposed solution to fill the same detection gaps. Improved threat detection requires deep learning-based methodologies and unsupervised techniques. We use GAN technology to detect cyber threats which occur in IoT-based network systems. The outcome produces a detection model with increased accuracy along with stronger reliability. This solution delivers better detection rates while requiring fewer accuracy terms and achieves better reliability. The ensemble learning classify compromised True Negative Rates as well as Hit Detection Rates when it comes to identifying Bruteforce attacks among others. The study results display uniformity throughout every dataset inquiry. System data maintains confidential integrity through utilization of this security principle. Telemetry-based threat spotting integrated with continuous monitoring runs opposed to traditional system patterns; this model implements it. The main objective of this work focuses on improving IoT-driven systems' resilience by means of increased efficiency with no rise in resource requirements. The conceptual framework with technology foundation are detailed throughout this chapter. The end objective aims to improve IoT systems resilience but it maintains operational efficiency and resource-intensive operation. The chapter presents details about the conceptual design along with technological bases and implications of running model applications. The terms include Cyber Security IDS Deep Learning IOT Security Generative Adversarial Network LSTM CNN Data Privacy Role Based Access and Secure Data Management Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 INTRODUCTION The IDS is a system which monitors the network for suspicious activities; it’s so configured that when these onlooked by it, the IDS immediately raises alarms. Because most system properties relate to different applications, then intrusion detection protects the protection line of a system working with benign traffic or normal flow patterns, as well as the set of the attack rules to determine harmful activity from non malicious movement. Data mining is applied to design and deploy an IDS in a strong way, on one hand, closer accuracy with the conventional IDS is provided and on the other hand, such applications are described for the contemporary, smart applications of cyber attacks. Growth of critical infra encompassing increasing organizations is observed, and with it the components within (IICs) manage an increasing number of connections of device used in setup of IIoT. Existing IDS are confronted with several severe challenges, in particular, when considered in terms of evolution in performance metrics. At the same time, all methods for evaluation metrics of IDS, apart from a majority of then, are quite broad. In this paper, an IDS is proposed and effective in utilizing deep auto encoder based LSTM model to bridge the gap of poor detection ratio and high false positive rates. Given the present urgency of developing real time IDS for IoT enabled infrastructures, this chapter concentrates on the development of a proactive, intelligent, IDS solution, capable to monitor ongoing activities, detect anomalies, and take actions in a proactive manner so as to mitigate threats in real time. With the help of lightweight and adaptive detection approaches, the proposed framework enables to secure IoT network without affecting its performance and scalability. This paper then examines the approach’s motivation working from the technical underpinnings of the solution, and how the solution could serve to reinforce cybersecurity in such an increasingly connected world. LITERATURE SURVEY Consequently, all of these ensemble techniques have seen an enormous surge forward in image classification: Alex-net, VGG, Inception, etc. Thus, these networks were considered for judging the relative knockouts in 'Imagenet' datasets. It then evaluates the effectiveness in classifying videos of Kinetics400 and UCF101 and whether this would suggest the possibility of success in video classification finally. These models are then compared with them models in which the margins of error are used to compare them for such purposes, and two of the networks that had the lowest margins were chosen to continue analyzing them. The video classification analysis consists of these ones. Since these networks have succeeded in this manner, inputs are taken into them through sensors that specify human performance in the video. Notably, we have well satisfied success rate of 70% on the "ResNet" and "Inception" benchmarks which indicates powerful performance and popularity of this idea. At the same time as there is a rapid increase in number of connected devices, the cloud data grows rapidly and the same time there is a rapid increase of complexity causing the data processing. IoT devices are often contain some sensitive information that could enable the industries or personal to make critical live time decision that has a real time effect. Most of the IoT devices are satisfying these constraints of high battery, memory shortage and low energy storing and in that case, they are posing against attacks as they lack sufficient resources to support security software, therefore an inherent risk is created on the IoT network. In order to tackle rising challenges with IoT devices and their expanding requirement. Edge computing platforms move some difficult computing tasks from devices to end edges servers and thus helps in mitigating the ‘complexity’ of processing tasks located in the device data. The majority of the existing related works focus on solving on live IoT device and data protection. Furthermore, we scan a number of papers desiring attention to numerous remedies that depend on MEC. The first part of the paper contains comprehensive study on modern intrusions and various security solutions. It analyzes the approach on the basis of MEC using various ensemble methods. Datasets, performance metrics and real world network intrusion deployment strategies are appeared in this paper. We conclude that MEC is used to tailor the NIDS networks. Health care prospects reside in IoT. There is a marked leap towards potential means of medical services regarding the Internet of Medical Things. The seeing of threats in health care connected devices puts the patient concerned as well as the privacy at risk if one services multitude merits. Because these devices are used to provide such large medical services to the large population within a network, it is necessary to develop strongly secured models ensuring privacy and safety of a patient within this network. Therefore, the aim of this paper is to propose a model of intrusion detection in networks with the help of Tree classifier prognosis. Besides, it has been designed with a view to, reduce the input data dimensional and hence, facilitates to handle anomalies rapidly and strongly to ensure a 94.23% accuracy rate. Deep learning’s capacity for intrusions and so anomaly recognition has been explored by several researchers. A specific intrusion detection system based on a tree classifier for the Internet of Medical Things was proposed for instance. It is stated that ensemble models can maximize the accuracy of unauthorized access attempt detection in sensitive network era. Agarwal and Gupta ( 2023 ) also make another significant contribution to making the blockchain-enabled authentication mechanism for IoT cybersecurity. METHODOLOGY a) Proposed Work : Here is a deep learning model that was introduced for identifying vulnerabilities and intrusions into real time systems. This framework involves unsupervised, deep learning discriminative approaches.It was introduced GAN setting RBN to act against threats targeting IIoT devices, performance is efficient in IDS frameworks are evaluated on industrial IoT internal and external networks under some datasets. b) Architecture of the System : Data flow diagram DFD is a vital tool used for modeling the functional aspects through system. It helps in illustrating the different components of systems impact, including internal processes, the data involved in processing, external entities which interacts with system, and the way data is flowing between elements. The main purpose is to show flow of information within system, and how it is transmitted at various stages. This diagrams are used as a graphical format to follow the flow of data from inputs to outputs. It is used as different levels of detailed view. By breaking it down into multiple layers, it becomes easier to understand the system functionality and the information is being travelled changes throughout. c) User-Signup : It is important that the new individuals on boarding new users into the system be secure and in-built within a User Sign-up Module. It collects important user identification like usernames, email addresses, and a password as secure as possible. Validation protocols are incorporated to keep the data consistent and intact as well as security procedures such as password complexity requirements and email verification. Use encryption techniques to store user credentials to prevent unauthorized access or data leakage. After successful registration, this information will be securely recorded into the back-end database of the system for authenticated personalized services usage. The security of the sign-up process by ensuring that the person attempting to create an account has access to the provided email or phone number. It’s an important tool in preventing unauthorized account creation and ensuring that only legitimate users can complete the registration process. d) User Sign In : The provided login credentials are verified against the database and if they match any data about a user, it makes the user registered for the application access. That basically describes it, a simple case of matching the username/email and password of the user, which is very securely protected via hash algorithm. Igor simulates failure, giving appropriate feedback if login fails, in some cases, repeated failures will result in temporary account lockout or CAPTCHA verification to prevent brute force attacks. Authentication will be taken and the user will gain access to a secured session where the core functions in the system can then be accessed. 1. Add Account The Home module is the one control panel all users can avail themselves of to do all that is important in the application. It also includes immediate access to major modules such as prediction, model performance, account settings, etc., with an overview of recent activities, and usually involves one or more user notifications or system alerts. Whereas home based design focuses on usability first; simple navigation, content prioritization, and the usage patterns based roles in creating a responsive design matter than the layout and framework. 2. Home This is a control panel where users have access to all the mandatory features of the application and it is called as Home module. They present users with the main modules (such as prediction, model performance, etc.), as well as quick access to them; also including notifications or system alerts. The first is home based design that puts the usability first and simple navigation, Content prioritization and adaptation of responsive layout according to the user roles and usage pattern makes it familiar. e) Blockchain Integration The purpose of the application, its evolution, and the contributors are detailed in the application’s About section. It provides enough information about the system’s mission, the system’s coverage, and the system’s value end. Links to the documentation, external references, support contact details and acknowledgements can be included in the About section as well. It gives transparency to the user and thus creates trust and professional sector credibility of the application in academic and commercial cases. f ) Notebook The Notebook module is a virtual space where structed or unstructured information may be documented, organized, and stored through different mediums in it. It might have text input, attachment of media files, time stamping, placing in categories for quick retrieval. This module also allows depending on system architecture for collaborative entries sometimes, submissions, their versions and exports. Usually, it builds a rich text editor on the online front end and keeps user input continuous between all user sessions, all devices. In this project, notebook uses three widely recognized datasets. CONCLUSION The challenges and hurdles in earlier works have been styled around by using deep learning for fast extermination and detection for threats. The paper further explains in detailed application for deep learning techniques in identifying malware-cyberattack disguised as expected behavior. Summarizing deep learning techniques that are generative models are developed. The precision results were gathered for the context of this study; the "provisioned" dictionaries are also relevant within the research field. The work has a demonstration by experimentation on IDS and Cyber security Cyber-attacks that are observable in successful collaborations of technological environments. This also looks like which of these perform better than the others in various DL techniques. To enhance deep learning methods are being explored to produce accuracy by classifying threat effectively. These approaches are aimed to match an exceed performance for future use. The efficiency of supervised approach is confirmed thorough training and testing IDS rapidly; therefore, the properties would incorporated with newly formed in time IDS for recognizing intruders with malicious behavior. FUTURE SCOPE Expansion in Internet of things (IoT) devices and its network of communication is ever growing, which renders the challenge of securing all the connected devices too still. There are numerous options for future work and enhancement when this project would research the possible scopes of application in the future into new features of deep learning architecture like transformer-based models or graph neural network models, which would allow identifying better complex relationships defined in the traffic flow of the most advanced metering infrastructures as these models show better performance for learning dynamic patterns that emerge from constantly changing IoT ecosystems. One further promising area of future work to be explored would be to explore federated and edge learning frameworks that enable training of models at the data source and without compromising on privacy. These approaches would particularly be advantageous in resource limited IoT devices with minimum possible delay and a decentralized detection. Declarations Author Contribution shaik.Ayazul Hasan and Barri.Vaishnavi, conceptualized the study and developed the initial methodology. Ch.Meghana was responsible for implementing the deep learning models and conducting experiments on the KDDCUP99 and UNSW-NB15 datasets. Dr. K. Krishna Jyothi. provided guidance on the research framework, supervised the overall project, and reviewed all analytical results. Ch.Meghana. and Barri.Vaishnavi wrote the main manuscript text. All authors reviewed, edited, and approved the final manuscript. References Deep learning underpins the philosophy, LeCun, Y., Bengio, Y., Hinton, G.: Nature. 521 (7553), 436–444 (2015). and technical advances that revise machine learning and artificial intelligence The deep learning, Krizhevsky, A., et al.: success stories about deep convolutional neural networks to classify the images from the ImageNet dataset are stated in this work. Commun. ACM. 60 (2), 84–90 (2017) Islam, A.M., Ali, M.S., Ali, M.M., Haque, M.F., Das, A.A., Hossain, M.M., Duranta, D.S.: and M.A. Rahman. The paper discusses a method to classify melanoma skin lesions using deep convolutional networks and transfer learning. In Proceedings of the 1st International Conference on Artificial Intelligence and Data Analytics (CAIDA) Ahmim, A., Derdour, M.A., Ferrag, M.A.: Int. J. Commun. Syst 31 (9): e3547. (2018) Ahmim, A., Maglaras, L., Ferrag, M.A., Derdour, M., Janicke, H.: The evolution of a hierarchical intrusion detection model for enhanced security based on decision trees and rule-based systems. In Proceedings of the 15th International Conference on Distributed Computing and Sensor Systems (DCOSS), 228–233, (2019) Dewa, Z., Maglaras, L.A.: This paper provides a review of the main data mining techniques employed for intrusion detection systems and highlights the key strategies and challenges faced. Int. J. Adv. Comput. Sci. Appl. 7 (1), 1–10 (2016) The authors proposed an, Stewart, B., Rosa, L., Maglaras, L.A., Cruz, T.J., Ferrag, M.A., Simoes, P., Janicke, H.: adaptive intrusion detection system for SCADA networks equipped to adjust with changing network topology. EAI Endorsed Trans. Industrial Networks Intell. Syst. 4 (10), e4 (2017) Ferrag, M.A., Maglaras, L., Moschoyiannis, S., Janicke, H.: A comparative review of the utilization of deep learning techniques in cyber-security via intrusion detection was provided by this paper, with consideration given to perspective datasets and approaches. J. Inform. Secur. Appl., 50 , 102419 Imrana, Y., Xiang, Y., Ali, L., Abdul-Rauf, Z.: In Expert Systems with Applications, 185, 115524, the authors propose bidirectional LSTM deep learning models for detection of intrusion with the aim of capturing both past and future network events Salih, A.A., Ameen, S.Y., Zeebaree, Z.R., Sadeeq, M.A., Kak, S.F., Omar, N., Ibrahim, I.M., Yasin, H.M., Rashid, Z.N.: Z. S. Ageed. Deep learning methods used in intrusion detection are surveyed in this paper, discussing several approaches and challenges. Asian J. Res. Comput. Sci., 9 , Issue 4, (2021). Pages 50 to 64. The practical application of the convolutional neural networks for image classification and other related activities are presented in this manuscript. A. Azevedo, F. Portela: Proceedings of the International Conference on (2022) Gupta, K., Sharma, D.K., Gupta, K.D., Kumar, A.: A Tree Classifier-Based Network Intrusion Detection Model for Internet of Medical Things. Comput. Electr. Eng., (2022) Kandhro, A., Al Enezi, S.M., Ali, F., Kehar, A., Fatima, K., Uddin, M., Karuppayah, S.: Detection of Real-Time Malicious Intrusions and Attacks in IoT Empowered Cybersecurity Infrastructures. IEEE Access., (2023) Agarwal, V., Gupta, P.: A Blockchain-Enabled Authentication Scheme for IoT Cybersecurity Infrastructure, in Proc. Int. Conf. on Applied Intelligence and Sustainable Computing (ICAISC), (2023) Kotha, S., Tekulapalli, P.R., Pogaku, S.S.V., Mohammed, G.B.: Real-time Detection of Malicious Intrusions and Attacks in Cybersecurity Infrastructures Enabled by IoT, MATEC Web of Conferences, (2024) Al Enezi, S.M., Kandhro, I.A., Ali, F.: Log-based Anomaly Detection using LSTM for Secure Systems. J. Cybersecur. Res., 11 , 2, (2022) Mohammed, G.B., Pogaku, S.S.V., Kotha, S.: Deep Learning for Securing Electronic Health Records: An Autoencoder Approach. Health Inf. J., 28 , 1, (2022) Zhang, Y., Wu, H., Li, M.: GAN-Based Synthetic Attack Generation for Deep Cybersecurity Models. Springer J. Inform. Secur. 17 (3), 223–234 (2021) Lee, D., Kim, J., Han, S.: Dynamic Access Control Using Deep Learning in Cloud Infrastructure, Cloud Security Transactions, vol. 6, pp. 45–52, (2023) Fatima, K., Uddin, M.: Transfer Learning for Cybersecurity: A Lightweight Approach to Threat Detection, Cyber AI Review, vol. 4, no. 2, pp. 89–98, (2023) Additional Declarations No competing interests reported. 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-6625113","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":456547808,"identity":"9cf10665-a608-43e4-9194-76191fa6c7b5","order_by":0,"name":"Ch. Meghana","email":"data:image/png;base64,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","orcid":"","institution":"Geethanjali college of engineering and technology","correspondingAuthor":true,"prefix":"","firstName":"Ch.","middleName":"","lastName":"Meghana","suffix":""},{"id":456547809,"identity":"f5d3227a-9ff2-4778-840b-1dc9bbb5d717","order_by":1,"name":"B. Vaishnavi","email":"","orcid":"","institution":"Geethanjali college of engineering and technology","correspondingAuthor":false,"prefix":"","firstName":"B.","middleName":"","lastName":"Vaishnavi","suffix":""},{"id":456547810,"identity":"b74cc9ae-7722-48eb-9662-375070220063","order_by":2,"name":"Shaik. Ayazul Hasan","email":"","orcid":"","institution":"Geethanjali college of engineering and technology","correspondingAuthor":false,"prefix":"","firstName":"Shaik.","middleName":"Ayazul","lastName":"Hasan","suffix":""}],"badges":[],"createdAt":"2025-05-09 05:08:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6625113/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6625113/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82851997,"identity":"1e56a63f-a28a-4ac3-8174-337192c41d80","added_by":"auto","created_at":"2025-05-16 03:32:59","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":53734,"visible":true,"origin":"","legend":"\u003cp\u003eArchitecture of the system\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/5880acceb1d6bbfdd5462d30.png"},{"id":82851686,"identity":"83cdc787-b653-4bf3-9e2b-05fb24f87178","added_by":"auto","created_at":"2025-05-16 03:24:59","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":30282,"visible":true,"origin":"","legend":"\u003cp\u003eAccuracy comparison grapgs-KDDCUP99 Dataset\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/de09aac88b761ed6f84cbe1c.png"},{"id":82851684,"identity":"2209e521-8d86-4058-9c2c-141858bbc4c3","added_by":"auto","created_at":"2025-05-16 03:24:59","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":32054,"visible":true,"origin":"","legend":"\u003cp\u003ePrecission comparison graphs-KDDCUP99 Datase\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/15340ad373cefda372bdf535.png"},{"id":82851999,"identity":"dceb9c30-171f-4ecc-a655-d323e5b6a0c9","added_by":"auto","created_at":"2025-05-16 03:32:59","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":30983,"visible":true,"origin":"","legend":"\u003cp\u003eRecall comparison graphs-KDDCUP99 Dataset\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/b2c5d484807aef74aace6ecf.png"},{"id":82851693,"identity":"42cd9dea-534a-4439-9e0f-5bf96c95cde8","added_by":"auto","created_at":"2025-05-16 03:24:59","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":28940,"visible":true,"origin":"","legend":"\u003cp\u003eAccuracy comparison graphs-UNSW-NB15 Dataset\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/0eda08860b60c7d49884bfa9.png"},{"id":82851691,"identity":"24ddc81a-2f00-43c4-be33-79946f67759c","added_by":"auto","created_at":"2025-05-16 03:24:59","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":27211,"visible":true,"origin":"","legend":"\u003cp\u003ePrecission comparison graphs-UNSW-NB15 Dataset\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/6ee3d37a50b0434e599357be.png"},{"id":82852582,"identity":"0cc805e7-e453-4b49-a573-7aa6c5822ba7","added_by":"auto","created_at":"2025-05-16 03:40:59","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":23499,"visible":true,"origin":"","legend":"\u003cp\u003eUnnumbered image in the Materials \u0026amp; Methods section.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/1f7e83acf87f0f18677b61be.png"},{"id":82851694,"identity":"a5554f0a-7474-4a7d-9d8d-e5e6816f16d9","added_by":"auto","created_at":"2025-05-16 03:24:59","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":408189,"visible":true,"origin":"","legend":"\u003cp\u003eUnnumbered image in the Materials \u0026amp; Methods section.\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/67b0fc696bf92fd8c72b3da3.png"},{"id":84031467,"identity":"6dd44749-66e0-41b3-ab15-ff3f1b1c9e5e","added_by":"auto","created_at":"2025-06-06 02:31:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":923639,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6625113/v1/face7894-1235-4a37-a3f0-39a7f636fb53.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Cyber Threat Using Deep Learning","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eThe IDS is a system which monitors the network for suspicious activities; it\u0026rsquo;s so configured that when these onlooked by it, the IDS immediately raises alarms. Because most system properties relate to different applications, then intrusion detection protects the protection line of a system working with benign traffic or normal flow patterns, as well as the set of the attack rules to determine harmful activity from non malicious movement. Data mining is applied to design and deploy an IDS in a strong way, on one hand, closer accuracy with the conventional IDS is provided and on the other hand, such applications are described for the contemporary, smart applications of cyber attacks. Growth of critical infra encompassing increasing organizations is observed, and with it the components within (IICs) manage an increasing number of connections of device used in setup of IIoT. Existing IDS are confronted with several severe challenges, in particular, when considered in terms of evolution in performance metrics. At the same time, all methods for evaluation metrics of IDS, apart from a majority of then, are quite broad. In this paper, an IDS is proposed and effective in utilizing deep auto encoder based LSTM model to bridge the gap of poor detection ratio and high false positive rates.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eGiven the present urgency of developing real time IDS for IoT enabled infrastructures, this chapter concentrates on the development of a proactive, intelligent, IDS solution, capable to monitor ongoing activities, detect anomalies, and take actions in a proactive manner so as to mitigate threats in real time. With the help of lightweight and adaptive detection approaches, the proposed framework enables to secure IoT network without affecting its performance and scalability. This paper then examines the approach\u0026rsquo;s motivation working from the technical underpinnings of the solution, and how the solution could serve to reinforce cybersecurity in such an increasingly connected world.\u003c/p\u003e"},{"header":"LITERATURE SURVEY","content":"\u003cp\u003eConsequently, all of these ensemble techniques have seen an enormous surge forward in image classification: Alex-net, VGG, Inception, etc. Thus, these networks were considered for judging the relative knockouts in 'Imagenet' datasets. It then evaluates the effectiveness in classifying videos of Kinetics400 and UCF101 and whether this would suggest the possibility of success in video classification finally. These models are then compared with them models in which the margins of error are used to compare them for such purposes, and two of the networks that had the lowest margins were chosen to continue analyzing them. The video classification analysis consists of these ones. Since these networks have succeeded in this manner, inputs are taken into them through sensors that specify human performance in the video. Notably, we have well satisfied success rate of 70% on the \"ResNet\" and \"Inception\" benchmarks which indicates powerful performance and popularity of this idea.\u003c/p\u003e \u003cp\u003eAt the same time as there is a rapid increase in number of connected devices, the cloud data grows rapidly and the same time there is a rapid increase of complexity causing the data processing. IoT devices are often contain some sensitive information that could enable the industries or personal to make critical live time decision that has a real time effect. Most of the IoT devices are satisfying these constraints of high battery, memory shortage and low energy storing and in that case, they are posing against attacks as they lack sufficient resources to support security software, therefore an inherent risk is created on the IoT network. In order to tackle rising challenges with IoT devices and their expanding requirement. Edge computing platforms move some difficult computing tasks from devices to end edges servers and thus helps in mitigating the \u0026lsquo;complexity\u0026rsquo; of processing tasks located in the device data. The majority of the existing related works focus on solving on live IoT device and data protection. Furthermore, we scan a number of papers desiring attention to numerous remedies that depend on MEC. The first part of the paper contains comprehensive study on modern intrusions and various security solutions. It analyzes the approach on the basis of MEC using various ensemble methods. Datasets, performance metrics and real world network intrusion deployment strategies are appeared in this paper. We conclude that MEC is used to tailor the NIDS networks.\u003c/p\u003e \u003cp\u003eHealth care prospects reside in IoT. There is a marked leap towards potential means of medical services regarding the Internet of Medical Things. The seeing of threats in health care connected devices puts the patient concerned as well as the privacy at risk if one services multitude merits. Because these devices are used to provide such large medical services to the large population within a network, it is necessary to develop strongly secured models ensuring privacy and safety of a patient within this network. Therefore, the aim of this paper is to propose a model of intrusion detection in networks with the help of Tree classifier prognosis. Besides, it has been designed with a view to, reduce the input data dimensional and hence, facilitates to handle anomalies rapidly and strongly to ensure a 94.23% accuracy rate.\u003c/p\u003e \u003cp\u003eDeep learning\u0026rsquo;s capacity for intrusions and so anomaly recognition has been explored by several researchers. A specific intrusion detection system based on a tree classifier for the Internet of Medical Things was proposed for instance. It is stated that ensemble models can maximize the accuracy of unauthorized access attempt detection in sensitive network era. Agarwal and Gupta (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) also make another significant contribution to making the blockchain-enabled authentication mechanism for IoT cybersecurity.\u003c/p\u003e"},{"header":"METHODOLOGY","content":"\u003cp\u003e \u003cb\u003ea) Proposed Work\u003c/b\u003e:\u003c/p\u003e \u003cp\u003eHere is a deep learning model that was introduced for identifying vulnerabilities and intrusions into real time systems. This framework involves unsupervised, deep learning discriminative approaches.It was introduced GAN setting RBN to act against threats targeting IIoT devices, performance is efficient in IDS frameworks are evaluated on industrial IoT internal and external networks under some datasets.\u003c/p\u003e \u003cp\u003e \u003cb\u003eb) Architecture of the System\u003c/b\u003e:\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eData flow diagram DFD is a vital tool used for modeling the functional aspects through system. It helps in illustrating the different components of systems impact, including internal processes, the data involved in processing, external entities which interacts with system, and the way data is flowing between elements. The main purpose is to show flow of information within system, and how it is transmitted at various stages. This diagrams are used as a graphical format to follow the flow of data from inputs to outputs. It is used as different levels of detailed view. By breaking it down into multiple layers, it becomes easier to understand the system functionality and the information is being travelled changes throughout.\u003c/p\u003e \u003cp\u003e \u003cb\u003ec) User-Signup\u003c/b\u003e:\u003c/p\u003e \u003cp\u003eIt is important that the new individuals on boarding new users into the system be secure and in-built within a User Sign-up Module. It collects important user identification like usernames, email addresses, and a password as secure as possible. Validation protocols are incorporated to keep the data consistent and intact as well as security procedures such as password complexity requirements and email verification. Use encryption techniques to store user credentials to prevent unauthorized access or data leakage. After successful registration, this information will be securely recorded into the back-end database of the system for authenticated personalized services usage. The security of the sign-up process by ensuring that the person attempting to create an account has access to the provided email or phone number. It\u0026rsquo;s an important tool in preventing unauthorized account creation and ensuring that only legitimate users can complete the registration process.\u003c/p\u003e \u003cp\u003e \u003cb\u003ed) User Sign In\u003c/b\u003e:\u003c/p\u003e \u003cp\u003eThe provided login credentials are verified against the database and if they match any data about a user, it makes the user registered for the application access. That basically describes it, a simple case of matching the username/email and password of the user, which is very securely protected via hash algorithm. Igor simulates failure, giving appropriate feedback if login fails, in some cases, repeated failures will result in temporary account lockout or CAPTCHA verification to prevent brute force attacks. Authentication will be taken and the user will gain access to a secured session where the core functions in the system can then be accessed.\u003c/p\u003e\n\u003ch3\u003e1. Add Account\u003c/h3\u003e\n\u003cp\u003eThe Home module is the one control panel all users can avail themselves of to do all that is important in the application. It also includes immediate access to major modules such as prediction, model performance, account settings, etc., with an overview of recent activities, and usually involves one or more user notifications or system alerts. Whereas home based design focuses on usability first; simple navigation, content prioritization, and the usage patterns based roles in creating a responsive design matter than the layout and framework.\u003c/p\u003e\n\u003ch3\u003e2. Home\u003c/h3\u003e\n\u003cp\u003eThis is a control panel where users have access to all the mandatory features of the application and it is called as Home module. They present users with the main modules (such as prediction, model performance, etc.), as well as quick access to them; also including notifications or system alerts. The first is home based design that puts the usability first and simple navigation, Content prioritization and adaptation of responsive layout according to the user roles and usage pattern makes it familiar.\u003c/p\u003e \u003cp\u003e \u003cb\u003ee) Blockchain Integration\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe purpose of the application, its evolution, and the contributors are detailed in the application\u0026rsquo;s About section. It provides enough information about the system\u0026rsquo;s mission, the system\u0026rsquo;s coverage, and the system\u0026rsquo;s value end. Links to the documentation, external references, support contact details and acknowledgements can be included in the About section as well. It gives transparency to the user and thus creates trust and professional sector credibility of the application in academic and commercial cases.\u003c/p\u003e \u003cp\u003ef\u003cb\u003e) Notebook\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThe Notebook module is a virtual space where structed or unstructured information may be documented, organized, and stored through different mediums in it. It might have text input, attachment of media files, time stamping, placing in categories for quick retrieval. This module also allows depending on system architecture for collaborative entries sometimes, submissions, their versions and exports. Usually, it builds a rich text editor on the online front end and keeps user input continuous between all user sessions, all devices. In this project, notebook uses three widely recognized datasets.\u003c/p\u003e"},{"header":"CONCLUSION","content":"\u003cp\u003eThe challenges and hurdles in earlier works have been styled around by using deep learning for fast extermination and detection for threats. The paper further explains in detailed application for deep learning techniques in identifying malware-cyberattack disguised as expected behavior. Summarizing deep learning techniques that are generative models are developed. The precision results were gathered for the context of this study; the \"provisioned\" dictionaries are also relevant within the research field. The work has a demonstration by experimentation on IDS and Cyber security Cyber-attacks that are observable in successful collaborations of technological environments. This also looks like which of these perform better than the others in various DL techniques. To enhance deep learning methods are being explored to produce accuracy by classifying threat effectively. These approaches are aimed to match an exceed performance for future use. The efficiency of supervised approach is confirmed thorough training and testing IDS rapidly; therefore, the properties would incorporated with newly formed in time IDS for recognizing intruders with malicious behavior.\u003c/p\u003e"},{"header":"FUTURE SCOPE","content":"\u003cp\u003eExpansion in Internet of things (IoT) devices and its network of communication is ever growing, which renders the challenge of securing all the connected devices too still. There are numerous options for future work and enhancement when this project would research the possible scopes of application in the future into new features of deep learning architecture like transformer-based models or graph neural network models, which would allow identifying better complex relationships defined in the traffic flow of the most advanced metering infrastructures as these models show better performance for learning dynamic patterns that emerge from constantly changing IoT ecosystems. One further promising area of future work to be explored would be to explore federated and edge learning frameworks that enable training of models at the data source and without compromising on privacy. These approaches would particularly be advantageous in resource limited IoT devices with minimum possible delay and a decentralized detection.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eshaik.Ayazul Hasan and Barri.Vaishnavi, conceptualized the study and developed the initial methodology. Ch.Meghana was responsible for implementing the deep learning models and conducting experiments on the KDDCUP99 and UNSW-NB15 datasets. Dr. K. Krishna Jyothi. provided guidance on the research framework, supervised the overall project, and reviewed all analytical results. Ch.Meghana. and Barri.Vaishnavi wrote the main manuscript text. All authors reviewed, edited, and approved the final manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eDeep learning underpins the philosophy, LeCun, Y., Bengio, Y., Hinton, G.: Nature. \u003cb\u003e521\u003c/b\u003e(7553), 436\u0026ndash;444 (2015). and technical advances that revise machine learning and artificial intelligence\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe deep learning, Krizhevsky, A., et al.: success stories about deep convolutional neural networks to classify the images from the ImageNet dataset are stated in this work. Commun. ACM. \u003cb\u003e60\u003c/b\u003e(2), 84\u0026ndash;90 (2017)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIslam, A.M., Ali, M.S., Ali, M.M., Haque, M.F., Das, A.A., Hossain, M.M., Duranta, D.S.: and M.A. Rahman. The paper discusses a method to classify melanoma skin lesions using deep convolutional networks and transfer learning. In Proceedings of the 1st International Conference on Artificial Intelligence and Data Analytics (CAIDA)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhmim, A., Derdour, M.A., Ferrag, M.A.: Int. J. Commun. Syst 31 (9): e3547. (2018)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhmim, A., Maglaras, L., Ferrag, M.A., Derdour, M., Janicke, H.: The evolution of a hierarchical intrusion detection model for enhanced security based on decision trees and rule-based systems. In Proceedings of the 15th International Conference on Distributed Computing and Sensor Systems (DCOSS), 228\u0026ndash;233, (2019)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDewa, Z., Maglaras, L.A.: This paper provides a review of the main data mining techniques employed for intrusion detection systems and highlights the key strategies and challenges faced. Int. J. Adv. Comput. Sci. Appl. \u003cb\u003e7\u003c/b\u003e(1), 1\u0026ndash;10 (2016)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe authors proposed an, Stewart, B., Rosa, L., Maglaras, L.A., Cruz, T.J., Ferrag, M.A., Simoes, P., Janicke, H.: adaptive intrusion detection system for SCADA networks equipped to adjust with changing network topology. EAI Endorsed Trans. Industrial Networks Intell. Syst. \u003cb\u003e4\u003c/b\u003e(10), e4 (2017)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerrag, M.A., Maglaras, L., Moschoyiannis, S., Janicke, H.: A comparative review of the utilization of deep learning techniques in cyber-security via intrusion detection was provided by this paper, with consideration given to perspective datasets and approaches. J. Inform. Secur. Appl., \u003cb\u003e50\u003c/b\u003e, 102419\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eImrana, Y., Xiang, Y., Ali, L., Abdul-Rauf, Z.: In Expert Systems with Applications, 185, 115524, the authors propose bidirectional LSTM deep learning models for detection of intrusion with the aim of capturing both past and future network events\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalih, A.A., Ameen, S.Y., Zeebaree, Z.R., Sadeeq, M.A., Kak, S.F., Omar, N., Ibrahim, I.M., Yasin, H.M., Rashid, Z.N.: Z. S. Ageed. Deep learning methods used in intrusion detection are surveyed in this paper, discussing several approaches and challenges. Asian J. Res. Comput. Sci., \u003cb\u003e9\u003c/b\u003e, Issue 4, (2021). Pages 50 to 64.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe practical application of the convolutional neural networks for image classification and other related activities are presented in this manuscript. A. Azevedo, F. Portela: Proceedings of the International Conference on (2022)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGupta, K., Sharma, D.K., Gupta, K.D., Kumar, A.: A Tree Classifier-Based Network Intrusion Detection Model for Internet of Medical Things. Comput. Electr. Eng., (2022)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKandhro, A., Al Enezi, S.M., Ali, F., Kehar, A., Fatima, K., Uddin, M., Karuppayah, S.: Detection of Real-Time Malicious Intrusions and Attacks in IoT Empowered Cybersecurity Infrastructures. IEEE Access., (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAgarwal, V., Gupta, P.: A Blockchain-Enabled Authentication Scheme for IoT Cybersecurity Infrastructure, in Proc. Int. Conf. on Applied Intelligence and Sustainable Computing (ICAISC), (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKotha, S., Tekulapalli, P.R., Pogaku, S.S.V., Mohammed, G.B.: Real-time Detection of Malicious Intrusions and Attacks in Cybersecurity Infrastructures Enabled by IoT, MATEC Web of Conferences, (2024)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAl Enezi, S.M., Kandhro, I.A., Ali, F.: Log-based Anomaly Detection using LSTM for Secure Systems. J. Cybersecur. Res., \u003cb\u003e11\u003c/b\u003e, 2, (2022)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMohammed, G.B., Pogaku, S.S.V., Kotha, S.: Deep Learning for Securing Electronic Health Records: An Autoencoder Approach. Health Inf. J., \u003cb\u003e28\u003c/b\u003e, 1, (2022)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, Y., Wu, H., Li, M.: GAN-Based Synthetic Attack Generation for Deep Cybersecurity Models. Springer J. Inform. Secur. \u003cb\u003e17\u003c/b\u003e(3), 223\u0026ndash;234 (2021)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, D., Kim, J., Han, S.: Dynamic Access Control Using Deep Learning in Cloud Infrastructure, Cloud Security Transactions, vol. 6, pp. 45\u0026ndash;52, (2023)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFatima, K., Uddin, M.: Transfer Learning for Cybersecurity: A Lightweight Approach to Threat Detection, Cyber AI Review, vol. 4, no. 2, pp. 89\u0026ndash;98, (2023)\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"The terms include Cyber Security, IDS, Deep Learning, IOT Security, Generative Adversarial Network, LSTM, CNN, Data Privacy, Role Based Access and Secure Data Management","lastPublishedDoi":"10.21203/rs.3.rs-6625113/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6625113/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe network faces dangerous incidents from both inside and outside sources that destroy computer systems. The array defense system features intrusion detection as one of its active elements which protects network security from undetected problems. Traditional intrusion systems face difficulties with detection precision and rate along with generating several false alerts that require extra resources to identify enlightened attack patterns. An ensemble-based breach detection serves as the proposed solution to fill the same detection gaps. Improved threat detection requires deep learning-based methodologies and unsupervised techniques. We use GAN technology to detect cyber threats which occur in IoT-based network systems. The outcome produces a detection model with increased accuracy along with stronger reliability. This solution delivers better detection rates while requiring fewer accuracy terms and achieves better reliability. The ensemble learning classify compromised True Negative Rates as well as Hit Detection Rates when it comes to identifying Bruteforce attacks among others. The study results display uniformity throughout every dataset inquiry. System data maintains confidential integrity through utilization of this security principle. Telemetry-based threat spotting integrated with continuous monitoring runs opposed to traditional system patterns; this model implements it. The main objective of this work focuses on improving IoT-driven systems' resilience by means of increased efficiency with no rise in resource requirements. The conceptual framework with technology foundation are detailed throughout this chapter. The end objective aims to improve IoT systems resilience but it maintains operational efficiency and resource-intensive operation. The chapter presents details about the conceptual design along with technological bases and implications of running model applications.\u003c/p\u003e","manuscriptTitle":"Cyber Threat Using Deep Learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-16 03:24:54","doi":"10.21203/rs.3.rs-6625113/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":"69617349-bb00-47d2-8dd4-a16dde71eeba","owner":[],"postedDate":"May 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-06-06T02:23:46+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-16 03:24:54","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6625113","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6625113","identity":"rs-6625113","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}