Ranking Abuse via Strategic Pairwise Data Perturbations

This paper investigates the vulnerabilities of Maximum Likelihood Estimation (MLE)-based ranking systems, such as the Bradley-Terry model, to strategic data manipulation. It introduces an Adaptive Subset Selection Attack (ASSA) that formulates the manipulation task as a constrained combinatorial optimization problem, demonstrating that even small perturbations can significantly alter rankings. The findings underscore the critical need for more robust aggregation methods in decision-making systems, as MLE-based rankings exhibit a sharp phase-transition behavior when subjected to adversarial perturbations.