Coalition-Resistant Peer Rating for Long-Term Confidentiality

Abstract

The outsourced storage of sensitive data requires long-term confidentiality guarantees. Proactive secret sharing in a distributed storage system provides such guarantees. However, some storage service providers lack in reliability or performance for proactive secret sharing to be viable, which can threaten data confidentiality. Data owners need guidance to select the best-performing storage service providers. Aggregated peer ratings with a mediator can provide such guidance. Nevertheless, providers may rate each other inaccurately to undermine competitors. This rational behaviour must be taken into account to devise performance scoring mechanisms generating accurate aggregate scores. The natural formalism to analyse the strategies of rational agents is game theory. In this paper, we introduce a game-theoretic model of the peer rating strategies of providers. Within this model, we first show that an unincentivised performance scoring mechanism results in providers reporting inaccurate ratings. We then introduce an incentivised performance scoring mechanism, modelled as an infinitely repeated game, that discourages inaccurate ratings. We prove that this mechanism leads to accurate ratings and thus to accurate performance scores for each provider, within a margin depending on coalition sizes.