Anomaly detection and analysis in blockchain systems

Abstract For a long time, anomaly detection is such a well topic. Its use in the banking industry has aided in the detection of questionable hacking activity. In the network of bitcoin, since all nodes are unlabeled, there is no proof that any particular transaction is the result of illegal activity, this thesis seeks to identify transactions that are unusual or suspicious. Finding abnormalities in the bitcoin transaction network is the main objective. We discuss anomaly identification in this paper with particular reference to the Bitcoin transaction network(BTN). In this instance, anomalies behaviors is a proxy for apprehensive activity, thus our objective is to find anomalies in the dataset in terms of their percentage. To achieve this, we use the feature selection method which is sequential forward feature selection along with three ML techniques, k-means clustering, isolation forest, and support vector machine (SVM) and got the highest accuracy of 98.2% in SVM as compared to all other methods.