Detecting and Reacting to Changes in Reputation Flows

Abstract

Abstract. In inter-enterprise collaboration, autonomic services from different organizations must independently determine which other services they can rely on. Reputation-based trust management in the Pilarcos open service ecosystem combines shared experience information on the actors' past behaviour and the decision context to estimate the risks of a collaboration. The trust decision process is semi-automatic, with selected decisions forwarded to a human user. A particularly interesting feature of the decision process is incongruity, that is, unexpected changes in service performance. In the classical example, a previously well-behaved service turns malicious to cash in its good reputation as illgained monetary profit. If the reputation system swiftly reacts to such changes, it protects its user more efficiently and deters misbehaviour. We present a new model for detecting and reacting to incongruities in a reputation-based trust management system. The model is based on the concept of reputation epochs, dividing an actor's reputation into periods of internally consistent behaviour. In contrast to earlier approaches, this model provides the necessary flexibility for the trust management system to adjust to constantly changing business situations.

Behavioural Evaluation of Reputation-Based Trust Systems

“…The Variable-cutoff policy, in turn, could be circumvented with a large number of low-credibility reports before the first defection. We have discussed prompt reaction to notable changes in behaviour in other work [13], and proposed other extensions to the example policies in the thesis [3, 6.3].…”