COLLABORATIVE LEARNING FRAMEWORK FOR CYBERATTACK MITIGATION IN BLOCKCHAIN ENVIRONMENTS

Abstract

The purpose of this paper is to examine infiltration assaults and then provide a new methodology for detecting cyberattacks at the network layer of blockchain networks, such as flooding of transactions and brute password attempts. In particular, we start by creating and setting up a blockchain network in our lab. This blockchain network will be used to provide actual traffic data (both normal and attack data) for our learning models and to conduct experiments in real time to assess how well our suggested intrusion detection system performs. As far as we are aware, this is the first dataset for blockchain network cyberattacks to be synthesised in a lab. Next, we provide a brand-new collaborative learning approach that enables effective implementation in the blockchain network for attack detection. The primary concept of the suggested learning model is to allow blockchain nodes to actively gather data, use the Deep Belief Network to learn from the data, and then communicate what they have learnt with other blockchain nodes in the network. In this manner, unlike traditional centralised learning techniques, we may not only benefit from the expertise of every node in the network but also avoid the requirement to collect all raw data for training at a single node. Such a framework may help prevent excessive network overhead/congestion and the danger of revealing the privacy of local data. Our suggested intrusion detection architecture can identify assaults with an accuracy of up to 98.6%, as shown by both rigorous simulations and real-time trials.