Convex Hull Watchdog: Mitigation of Malicious Nodes in Tree-Based P2P Monitoring Systems

Abstract

Monitoring the global state in peer-to-peer networks through decentralized mechanisms allows targeted optimization and improvement of the peer-to-peer network. However, malicious nodes could aim to distort the process of gathering the global state through monitoring. In this paper we propose DOMiNo, a security solution for tree-based peer-to-peer monitoring mechanisms. It passively listens to incoming events, e.g. data, and rates its suspiciousness based on outlier detection, structural verification and sanity check mechanisms. For our main objective, which is to limit the monitoring error of the desired global view, we performed an extensive evaluation. Evaluation shows tolerance with normal fluctuations but effective filtering of outliers, that severely influence the global view. As our watchdog solution operates passively, we do not add any costs nor create new surface for attacks to the monitoring system.