Saving Stochastic Bandits from Poisoning Attacks via Limited Data Verification

This paper studies bandit algorithms under data poisoning attacks in a bounded reward setting. We consider a strong attacker model in which the attacker can observe both the selected actions and their corresponding rewards, and can contaminate the rewards with additive noise. We show that any bandit algorithm with regret O(log T ) can be forced to suffer a regret Ω(T ) with an expected amount of contamination O(log T ). This amount of contamination is also necessary, as we prove that there exists an O(log T ) regret bandit algorithm, specifically the classical Upper Confidence Bound (UCB), that requires Ω(log T ) amount of contamination to suffer regret Ω(T ). To combat such poisoning attacks, our second main contribution is to propose verification based mechanisms, which use limited verification to access a limited number of uncontaminated rewards. In particular, for the case of unlimited verifications, we show that with O(log T ) expected number of verifications, a simple modified version of the Explore-then-Commit type bandit algorithm can restore the order optimal O(log T ) regret irrespective of the amount of contamination used by the attacker. We also provide a UCB-like verification scheme, called Secure-UCB, that also enjoys full recovery from any attacks, also with O(log T ) expected number of verifications. To derive a matching lower bound on the number of verifications, we also prove that for any order-optimal bandit algorithm, this number of verifications Ω(log T ) is necessary to recover the order-optimal regret. On the other hand, when the number of verifications is bounded above by a budget B, we propose a novel algorithm, Secure-BARBAR, which provably achieves Õ(minC, T / √ B) regret with high probability against weak attackers (i.e., attackers who have to place the contamination before seeing the actual pulls of the bandit algorithm), where C is the total amount of contamination by the attacker, which breaks the known Ω(C) lower bound of the non-verified setting if C is large.