Comparative Study Analysis of MachineLearning Algorithms for Anomaly Detection in Blockchain

Abstract

Anomaly detection is one of the challenging problems encountered by the modern network security industry. In these last years, Blockchain technologies have been widely used in several application fields to improve data privacy and trustworthiness and security of systems. Despite being an effective tool, the blockchain is not impervious to cyberattacks. For instance, a successful 51% attack on Ethereum Classic exposed security flaws in the technology. Attacks can be viewed from a statistical standpoint as an aberrant finding that strongly deviates from the norm. Machine learning is a science whose objective is to discover insights, trends, and anomalies in massive data sets; as a result, it can be used to detect blockchain attacks. In this work, we define a federated learning-based anomaly detection system that is trained using aggregate data gathered from observing blockchain activity on the end device itself. Experiments on the whole historical logs of the Ethereum Classic network demonstrate our model’s ability to accurately identify assaults that have been made public while also automatically signing digital transactions for further protection. Therefore, it is necessary to create an anomaly detection system that can monitor networks for any dangerous actions and produce findings for the management authority in the end device itself. Several classification techniques and machine learning algorithms have been taken into consideration in our suggested article to categorize the accurate model.