Detection of Man-in-the-Middle Attacks in Model-Free Reinforcement Learning

This paper proposes a Bellman Deviation algorithm for the detection of man-in-the-middle (MITM) attacks occurring when an agent controls a Markov Decision Process (MDP) system using  model-free reinforcement learning. We show an intuitive necessary and sufficient ``informational advantage" condition  for  the proposed algorithm to guarantee the detection of attacks  with high probability, while  also avoiding false alarms.