An outlier mining-based malicious node detection model for hybrid P2P networks

With the increases of P2P applications and their users, the malicious attacks also increased significantly, which negatively impacts on the availability of the P2P networks and their users’ experience. This paper presents an outlier mining-based malicious node detection model for hybrid P2P networks. We first extract the local nodes’ frequent patterns from the nodes’ behavior patterns in subnets using the frequent behavior pattern mining approach, and then we produce and update the nodes’ global frequent behavior patterns by incrementally propagating and aggregating the local frequent behavior patterns. Finally, we identify outliers (i.e. the malicious nodes) using the local frequent behavior patterns and the global frequent behavior patterns. We also discuss how to recognize the different types of malicious nodes from outliers. Simulation results show that our strategy could detect malicious nodes with low false positive rate and low false negative rate.