Using Logistic Trust for Event Learning and Misbehaviour Detection

The advancement in communication technologies has enabled ad hoc networks to collect large volumes of information. This information is vulnerable to various types of attacks amongst which false information dissemination and on-off attacks offer biggest threats to the networks. As the data in ad hoc networks depends on the events, it is necessary for any detection mechanism to first determine the true events. Then the information about these events can be used to judge the behavior of the senders. Therefore, in this work, the correct event is first learned using information from different sources including the observations of the receiver itself. This information is later used to learn the behavior of the senders. The learned behavior combined with the opinions of the neighbors about the sender allows the detection of malicious and honest nodes. In this work, a logistic trust model is used to combine the observed behavior and opinions. It is observed that logistic trust results in a high accuracy of over 99% and a low error of less than 1% even when the events are changing rapidly. It is also shown that the scheme can detect malicious majority and identify true events with high probability.