Distributed Trust Management in Autonomic Networks
Doctoral Dissertation, Date: October 2007, Advisor: John S. Baras
The management of autonomic networks has gained more and more attentions because of their wide applications and control diffculties. Autonomic networks are decentralized and self-organized. Without global knowledge on the states of autonomic networks, it is diffcult to predict behaviors of such networks and thus to conduct proper network management and control. This dissertation is the starting point of my effort to theoretically understand the complex characteristics of autonomic networks. In particular, I focus on a specific application: distributed trust management.
We view trust among users as a set of relations established on the basis of trust credentials and required by specified policies. Two important components of a distributed trust management system are studied in this work: trust credential distribution and trust evaluation. In autonomic networks, trust credentials are distributed throughout the network. Given the mobility and dynamics of the networks, it is important to properly distribute trust credentials such that users are able to effciently obtain required credentials and update existing credentials. I present a trust credential distribution scheme based on network coding. After obtaining credentials in need, policies are required for users to evaluate trustworthiness of targets in a distributed way. In this dissertation, I model distributed trust evaluation as an estimation problem and trust evaluation policies based on local interactions are studied. I investigate the convergence of both deterministic and stochastic voting rules and prove their effectiveness with the present of misbehaving users.
Autonomic networks rely on collaboration among users. The conflict between the benefit from collaboration and the required cost for collaboration naturally leads to game-theoretic studies. I study collaboration based on cooperative games with communication constraints and give the conditions under which users are willing to collaborate. The results in this dissertation show that a well-designed trust management system is helpful to enforce collaboration. Besides collaboration, I show that trust can be used to the utility optimization problems as well. The effect of trust values is that in the routing and scheduling problems the trustworthiness of the node will be automatically considered and used. For example, packets will not be routed frequently to suspicious nodes.