A Privacy-enhanced Reputation System for Mobile Ad hoc Services

Abstract

MANET has been widely used for distributed communications. Various services can be developed based on MANET. However, lack of trust retards the success of many MANET based services (i.e., MANET services). This thesis work proposes AdRep -- a reputation system that could apply on various MANET services with privacy concern. We summarize the contributions of the thesis as below.First, we design AdRep based on a hybrid trust/reputation management framework that supports trust evaluation and reputation generation on any entities involved in the MANET based services. We adopted the trusted server to support more accurate and efficient trust/reputation evaluation and user privacy preservation. We designed a hybrid trust model that can support trust/reputation management in both MANET node devices and the trusted server. Due to the diversity of different MANET services, we designed a context aware trust model that can be adaptively configured based on different service contexts. Our trust model can be applied to implement various reputation systems for different MANET services, by providing concrete instances to the primitive trust model. Second, we implemented two concrete reputation systems AdChatRep and AdContRep to illustrate our framework and model's applicability. In AdChatRep, the context module is instantiated by MANET chatting, while the context module of AdContRep is instantiated by MANET content recommendation. We implemented AdChatRep based on Nokia N900 and AdContRep on Nokia N810. Both prototype systems have been demonstrated inside Nokia with sound feedback.Third, we further evaluated both AdChatRep and AdContRep using different methodologies. We launched a survey exploring usefulness, chatting trust influencing factors and reputation visualization UI design of AdChatRep before implementation. After prototyping it, we further conducted a user study based on AdChatRep usage and survey users' opinions on the usefulness, effectiveness and UI acceptance of AdChatRep. In addition, we also studied user behaviors based on real chatting logs and system records. We achieved sound feedback on AdChatRep from user study participants. For AdContRep, we evaluated its trust and reputation algorithms through Matlab simulations. We proved the effectiveness of our algorithms and their robustness against unfair rating attacks and on-off attacks.