Heuristic Discovery of Role-Based Trust Chains in Peer-to-Peer Networks
IEEE Transactions on Parallel and Distributed Systems
Credential chains are needed in trusted peer-to-peer (P2P) applications, where trust delegation must be established between each pair of peers at specific role level. Role-based trust is refined from the coarsegrained trust model used in most P2P reputation systems. This paper offers a novel heuristic-weighting approach to selecting the most likely path to construct a role-based trust chain. We apply history-sensitive heuristics to measure the path complexity and to assess the chaining
... y. Our method discovers successive edges of a trust chain, adaptively, to match with the demands in any given P2P application. New heuristic chaining algorithms are developed for backward, forward, and bi-directional discovery of trust chains. Our heuristic chain discovery scheme shortens the search time, reduces the memory requirement, and enhances the chaining accuracy in scalable P2P networks. Consider a trust graph over N credentials and M distinct role nodes. Our heuristic trust-chain discovery algorithms require O (N 2 logN) search time and O(M) memory space, if the secondary heuristics are generated off-line in advance. These are improved from O(N 3 ) search time and O(NM) space required in non-heuristic discovery algorithms developed by Li, Winsborough, and Mitchell (2003). Our analytical results are verified by extensive simulation experiments over typical classes of role-based trust graphs. Index Terms-Peer-to-peer networks, trust delegation, role-based credentials, heuristic semantics, and Internet applications -------------------- Credential Generation Process Preferred Roles Class-1: A.r ← B Randomly select B and select A.r using the small-world network model Specific roles with lower role-in-degree but larger entity-indegree. Class-2: A.r ← B.s Randomly select B.s and select A.r using the small-world network model Prefer A.r with larger role-in-degree but smaller entity-indegree. Class-3 : A.r ← A.r 1 .r 2 Randomly select A.r using the small-world network model. Randomly select r 1 and r 2 to yield A.r 1 .r 2 . Prefer roles with larger in-degree and then select r 1 and r 2 from a local role set randomly. Class-4 : A.r ← B.r 1 .r 2 . .. .r k Randomly select A.r using the small-world network model. Select r i for i = 1, 2, .., k from the role nodes by frequency of use.