6533b85bfe1ef96bd12bb42d

RESEARCH PRODUCT

Decentralised trust-management inspired by ant pheromones

subject

description

Computational trust is increasingly utilised to select interaction partners in open technical systems consisting of heterogeneous, autonomous agents. Current approaches rely on centralised elements for managing trust ratings (i.e. control and provide access to aggregated ratings). Consider a grid computing application as illustrating example: agents share their computing resources and cooperate in terms of processing computing jobs. These agents are free to join and leave, and they decide on their own with whom to interact. The impact of malicious or uncooperative agents can be countered by only cooperating with agents that have shown to be benevolent: trust relationships are established. Typically, this requires a centralised data-base storing information about past interactions and their outcome. In this article, we propose a novel, decentralised trust mechanism inspired by the nestmate recognition system in ants. More precisely, the concept of recognition pheromones, which stick to the agents and cannot be removed or counterfeit, is turned into algorithmic logic and interaction protocols. We demonstrate the potential benefit by using simulations of the grid scenario.