TrustNet: A trust and reputation management system in distributed environments

Abstract

With emerging Internet-scale open content and resource sharing, social networks, and complex cyber-physical systems, trust issues become prominent. Despite their rigorous foundations, conventional network security theories and mechanisms are inadequate at addressing such loosely-defined security issues in decentralized open environments. In this dissertation, we propose a trust and reputation management system architecture and protocols (TrustNet), aimed to define and promote trust as a first-class system parameter on par with communication, computation, and storage performance metrics. To achieve such a breakthrough, we need a fundamentally new design paradigm to seamlessly integrate trust into system design. Our TrustNet initiative represents a bold effort to approach this ultimate goal.

TrustNet is built on the top of underlying P2P and mobile ad hoc network layer and provides trust services to higher level applications and middleware. Following the TrustNet architecture, we design, implement, and analyze trust rating, trust aggregation, and trust management strategies. Especially, we propose three trust dissemination protocols and algorithms to meet the urgent needs and explicitly define and formulate end-to-end trust. We formulate trust management problems and propose the H-Trust, VectorTrust, and cTrust scheme to handle trust establishment and aggregation issues. We model trust relations as a trust graph in distributed environment to enhance accuracy and efficiency of trust establishment among peers. Leveraging the distributed Bellman-Ford algorithm, stochastic Markov chain process and H-Index algorithm for fast and lightweight aggregation of trust scores, our scheme are decentralized and self-configurable trust aggregation schemes.

To evaluate TrustNet management strategies, we simulated our proposed protocols in both unstructured P2P network and mobile ad hoc network to analyze and simulate trust relationships. We use software generated data as well as real world datasets. Particularly, the student contact patterns on the NUS campus is used as our trust communication model. The simulation results demonstrate the features of trust relationship dissemination in real environments and the efficiency, accuracy, scalability and robustness of the TrustNet system.