User Behavior Identification for Trust Management in Pervasive Computing Systems

Pervasive computing systems are becoming widespread as the next generation of computing systems. In pervasive computing systems, the users are frequently hopping from one computing space to another, which makes user profile management an essential component of the system. Because such a dynamic-populated computing system has to be new corner-friendly, malicious users with bad reputation to a system may take advantage of the system by claiming themselves as new users. In addition, user identity can also be compromised by controlling the victim node via normal invasions or even by inside identity thefts. A valid user identity is the foundation and prerequisite of trust. Failures to correctly identify a user are catastrophic to trust management system. Traditional intrusion detection systems can only detect an attack when it already happens. But before the intrusion the attackers may have some specific behavioral signatures indicating of their suspicious identities. We need a more powerful and practical system to identify these suspicious behaviors and signal an early warning to stop them. In this paper we assume that user identity is compromised either by attacking and controlling a node or by malicious users' claiming themselves as new users. We introduced a novel user behavior identification and profile management system to detect the suspicious behaviors, and a dynamically triggered reaction module to monitor the suspicious node and collect the necessary evidences