A Game Theory Based Multi Layered Intrusion Detection Framework for Wireless Sensor Networks

Abstract

Most of the existing intrusion detection frameworks proposed for wireless sensor networks (WSNs) are computation and energy intensive, which adversely affect the overall lifetime of the WSNs. In addition, some of these frameworks generate a significant volume of IDS traffic, which can cause congestion in bandwidth constrained WSNs. In this paper, we aim to address these issues by proposing a game theory based multi layered intrusion detection framework for WSNs. The proposed framework uses a combination of specification rules and a lightweight neural network based anomaly detection module to identify the malicious sensor nodes. Additionally, the framework models the interaction between the IDS and the sensor node being monitored as a two player non-cooperative Bayesian game. This allows the IDS to adopt probabilistic monitoring strategies based on the Bayesian Nash Equilibrium of the game and thereby, reduce the volume of IDS traffic introduced into the sensor network. The framework also proposes two different reputation update and expulsion mechanisms to enforce cooperation and discourage malicious behavior among monitoring nodes. These mechanisms are based on two different methodologies namely, Shapley Value and Vickery–Clark–Grooves (VCG) mechanism. The complexity analysis of the proposed reputation update and expulsion mechanisms have been carried out and are shown to be linear in terms of the input sizes of the mechanisms. Simulation results show that the proposed framework achieves higher accuracy and detection rate across wide range of attacks, while at the same time minimizes the overall energy consumption and volume of IDS traffic in the WSN.