Trust-based schemes are promising techniques to tackle inside attacks in distributed self-organized networks, such as mobile ad hoc networks and vehicular ad hoc networks. For the outside attackers, access control, authorization and authentication by cryptography can effectively thwart most of them. For the inside attackers, prevention based schemes such as cryptographic techniques are usually powerless. In the trust management system, trust is defined as the degree of belief that an entity can behave correctly in an observer’s perspective. Compared to prevention-based schemes, detection-based schemes, such as trust management, dynamically estimate the internal nodes behavior. Based on the results of the estimation, the detection system makes the decision whether the node is a malicious attacker. These detection-based schemes introduce a large amount of uncertainties due to the unpredictable behavior of each node in the networks. Therefore, in the trust management system, accurate trust assessment is playing a key role in the trust management. It is significantly affected by uncertainty. In order to obtain accurate trust of each entity in the network, we apply uncertain reasoning, coming from the artificial intelligence field, to trust management in the emerging networking paradigms.
In this dissertation, we focus on trust management with the probability methodologies. This is because that Bayesian probability can formulate the trust in distributed self-organized networks better than rule-based schemes. With the three methodologies, Bayesian inference, DST and Bayesian network model, in the uncertain reasoning, the trust formulated by them can effectively mitigate the inside attacks in those ad hoc networking paradigms.