Lightweight and distributed attack detection scheme in mobile ad hoc networks

Many of the widely used intrusion detection schemes, such as Self Organizing Map and Artificial Immune System, require heavy computational power in order to provide highly accurate results. These schemes have been successfully deployed in wired networks, which have high computational and bandwidth capabilities. Mobile ad hoc networks, however have limited resources and hence deploying such schemes are impractical. We propose a lightweight, low-computation, distributed intrusion detection scheme for mobile ad hoc networks termed the Distributed Hierarchical Graph Neuron (DHGN). The DHGN-based network is a new form of neural network, which consists of a hierarchical graph-based representation of input patterns. This pattern recognition scheme adopts a divide-and-distribute approach that divides an input pattern into a number of subpatterns, which are then concurrently processed for recognition. The first section of this paper provides an in-depth study of mobile ad hoc networks and current intrusion detection implementation in these networks. The second section of the paper provides an overview of the proposed two-stage cooperative intrusion detection system architecture and compares the proposed algorithm with Self Organizing Map classifier. The experiments show that our low computational scheme produced similar classification accuracy results to Self Organizing Map algorithm.