Inception: Efficiently Computable Misinformation Attacks on Markov Games

arXiv:2406.17114v1 Announce Type: cross
Abstract: We study security threats to Markov games due to information asymmetry and misinformation. We consider an attacker player who can spread misinformation about its reward function to influence the robust victim player's behavior. Given a fixed fake reward function, we derive the victim's policy under worst-case rationality and present polynomial-time algorithms to compute the attacker's optimal worst-case policy based on linear programming and backward induction. Then, we provide an efficient inception ("planting an idea in …

arxiv attacker attacks behavior can case cs.cr cs.gt cs.lg fake function games inception influence information misinformation player policy reward security security threats study threats under victim