A Systematic Security Approach in Wireless Mesh Networks

Abstract

Wireless mesh networking has emerged as a key technology to provide wide-coverage broadband networking. It benefits both service providers with low cost in network deployment, and end users with ubiquitous access to the Internet from anywhere at anytime. Wireless mesh networks are vulnerable to malicious attacks due to the nature of wireless communication and the lack of centralized network infrastructure. Meanwhile, the capacity of multi-radio multi-channel communication, the need for heterogeneous network integration, and the demand for multi-hop wireless communication often make traditional security mechanisms inefficient or infeasible. Therefore, wireless mesh networks pose new challenges and call for more effective and applicable solutions. In this work, we identify the requirement for a systematic security framework to protect wireless mesh networks and provide a security system with heterogeneity-aware intrusion prevention mechanism, cross-layer based intrusion detection technique, and a generic intrusion response model. Our major contributions lie in the following: (1) We identify the architecture heterogeneity of wireless mesh networks and proposed a novel heterogeneity-aware group key management framework which combines the logical key hierarchical technique together with the localized threshold-based technique. (2) To leverage link-aware routing characteristics, we present a cross-layer based anomaly detection model which utilizes machine learning algorithms for profile training and intrusion detection. (3) We address the automatic intrusion response problem in wireless mesh network by providing a generic response model to describe the dependency of system services and resources. The dependency graph is later used for damage cost assessment and response cost evaluation. (4) We build a wireless mesh network testbed and implemented a system prototype for intrusion detection system. Our simulation and experiment results show that our solutions outperform existing ones and are practical for wireless mesh networks in terms of communication overhead and performance speed.