Differentially Private Reinforcement Learning with Self-Play

arXiv:2404.07559v1 Announce Type: cross
Abstract: We study the problem of multi-agent reinforcement learning (multi-agent RL) with differential privacy (DP) constraints. This is well-motivated by various real-world applications involving sensitive data, where it is critical to protect users' private information. We first extend the definitions of Joint DP (JDP) and Local DP (LDP) to two-player zero-sum episodic Markov Games, where both definitions ensure trajectory-wise privacy protection. Then we design a provably efficient algorithm based on optimistic Nash value iteration and privatization …

agent applications arxiv constraints critical cs.ai cs.cr cs.lg cs.ma data definitions differential privacy information local play player privacy private problem protect real sensitive sensitive data stat.ml study world