Capturing Reputation Features in Multiagent Systems through Emerging Patterns

WALTER GRANDINETTI; CARLOS IVÁN CHESÑEVAR

Río Cuarto, Argentina

Congreso; 7MO Workshop Argentino de Investigadores en Ciencias de la Computacion (WICC); 2005

Multiagent systems and online communities rely on rating systems to inferthe reputation given to an individual within a particular context.The notion of reputation is essential for helping a given individual totrust in other individuals and for being himself reliable to others.Current techniques for computing individual´s reputations are solelybased on recent activities, facilitating a variety of possible attacks.Moreover, the amount of trust each agent has for a given contextis based just on his or her reputation.In this paper we outline a new way to thwart reputation-based attacks andto detect trends in behavioral patterns based on historical databy means ofknowledge discovery techniques, particularly those existing foremerging patterns.