Computational trust and rating for strategic service exchange under virtual anonymity, collusion, and blind partner selection

Computational Trust and Rating (CTR) schemes support decision-making in distributed multi-agent environments by aggregating ratings of past behavior into trustworthiness levels ( s) that influence agents’ utilities. Existing models rarely capture the closed-loop dependence between behavior, s, and utilities through explicit dynamics, nor do they provide analytical characterization of resilience against adversarial behavior. We propose a CTR scheme for Strategic Service Exchange (CTR-SSE), a framework based on pairwise agent interactions. CTR-SSE incorporates generic skimping and slandering attacks by subversive agents against honest agents and considers a challenging setting with virtual anonymity, -agnostic service provider selection, and uncertainty in service reception. To our knowledge, no prior analytical framework captures this combination of assumptions. For a class of service provision and rating (SPR) strategies, we derive closed-loop Markovian dynamics and their mean-value approximation to obtain steady-state s and utilities. A Stackelberg game formulation yields insights into resilience against subversion (subversive agents acquiring higher s) and DoS (some agents receiving no service). We quantify several intuitions of CTR scheme design and identify related structural properties, including the roles of the rating aggregation algorithm, SPR, indirect reciprocity, service uncertainty, the proportion of subversive agents, and their collusion. Numerical experiments and simulations validate the analysis.