A Probabilistic Framework for Distributed Localization of Attackers in MANETs

摘要

Mobile Ad-hoc Networks (MANETs) are frequently exposed to a wide range of cyber threats due to their unique characteristics. The lack of a centralized monitoring and management infrastructure and the dynamic nature of their topology pose new and interesting challenges for the design of effective security mechanisms. While conventional methods primarily focus on detecting attacks, in this work we focus on estimating the attackers' physical location in the network, and propose a probabilistic framework for aggregating information gathered from nodes reporting malicious activity in their vicinity. In order to overcome the limitations of the decentralized nature of MANETs, we present a distributed approach to attacker localization based on dynamically partitioning the network into clusters. These self-organizing clusters can (i) independently find the approximate location of the attackers in real time, and (ii) deploy trusted resources to capture attackers. We show, through experiments in a simulated environment, that our approach is effective and efficient.