A Statistical Signal Processing Framework for Secure Wireless and Sensor Networking

Cornell University, Ithaca NY

Investigators

Abstract

Wireless networks, sensor networks in particular, are vulnerable to intrusion and attack. Even the strongest encryption and authentication techniques are not adequate to provide security and trust. Because signals propagate in shared media, the very acts of transmission reveal crucial aspects of networking. More challenging for wireless sensor networks is that sensors are not physically secure; An adversary can then gain full control of a fraction of the sensors and program them to attack from within. This research investigates the theory and practical techniques of secure wireless and sensor networking. Statistical signal processing and machine learning techniques are developed to extract networking information. Countermeasures against traffic analysis are developed using network coding and randomized mixing techniques. Tradeoffs between stable throughput and secrecy (with delay constraints) are characterized theoretically. For sensor networks performing signal processing tasks, optimal attacking strategies using Byzantine sensors are investigated. Countermeasures are developed using robust and universal detection and estimation techniques. Collaborative sensor networking strategies are also investigated to counter Byzantine attacks.