Research On Reputation Scheme Based Malicious Nodes Detection For Mobile Ad Hoc Networks

A Mobile Ad hoc Network (MANET) consists of a collection of mobile nodes, which can self-organize into a multi-hop wireless network. MANET is now playing an important role in military communication, civil communication and temporary communication. It has the characteristics of no pre-existing infrastructure and dynamic changing topology, which make it more vulnerable to attacks than traditional wired networks. Usually security in MANETs was achieved by preventive schemes, which actually can not prevent inner malicious nodes from acting uncooperatively. Moreover, several selfish nodes may not be willing to fully cooperate for the purpose of saving battery resources. Aimed at the malicious behavior that would dramatically affect the overall performance and the security of MANETs, this paper propose a novel reputation model using subjective logic to defend against the attacks of inner nodes, and ensure the security of MANETs.Deep research has been performed in this paper, what is more, the existing detection methods of malicious nodes including those based on reputation scheme have been contrasted and analyzed. An Opinion Distance based Reputation Evaluation (ODRE) model was proposed in this paper in order to detect malicious nodes efficiently even when they colluded to interfere with the reputation assistant mechanism. In this model, the nodes were quantified by the improved Subject Logic in Ad Hoc method (SLAH). The local opinion was calculated and updated by mapping from communication evidence. ODRE model obtained the recommendation opinion with the aid of reputation sharing mechanism and combined it with the local opinion. The concept of opinion distance was introduced to filter malicious recommendations so that the false accusations laid by malicious nodes would have little influence on the performance of the reputation model. Eventually, processing the combination opinion yielded the final trustworthiness of a certain node, according to which other nodes would decide whether to interact with it. In this way, malicious nodes were isolated.Simulations were conducted to evaluate the proposed model on its performance. The results indicate that ODRE model performs much better than the previous similar models in detecting and isolating malicious nodes, especially when malicious colluding behaviors exist in the MANETs. Thus, this model is more robust.