A Self-organized Agent-based architecture for Power-aware Intrusion Detection in wireless ad-hoc networks

In this paper we propose SAPID-A Self-organized Agent-based architecture for Power-aware Intrusion Detection in wireless ad-hoc networks. We utilize an agent-based architecture that conserves the available bandwidth and segregates SAPID into two phases. Based on a power level metric and a hybrid metric that determine the duration and kinds of traffic that can be supported by a network-monitoring node, potential ad-hoc hosts are identified by repetitive training using an Adaptive Resonance Theory module. The agent architecture primarily consists of a Kohonen Self-Organizing Map to identify the appropriate patterns and recognize anomalies by way of unauthorized users in the network. A UNIX based session information file is utilized for testing and reporting possible intrusion attempts to the decision and action modules. Comprehensive experiments were carried out to clearly delineate and analyze the performance of the architecture.