Gradient Inversion in Federated Reinforcement Learning

Federated reinforcement learning (FRL) enables distributed learning of optimal policies while preserving local data privacy through gradientthis http URL, FRL faces the risk of data privacy leaks, where attackers exploit shared gradients to reconstruct local trainingthis http URLto traditional supervised federated learning, successful reconstruction in FRL requires the generated data not only to match the shared gradients but also to align with real transition dynamics of the environment (i.e., aligning with the real data transition distribution).To address this issue, we propose a novel attack method called Regularization Gradient Inversion Attack (RGIA), which enforces prior-knowledge-based regularization on states, rewards, and transition dynamics during the optimization process to ensure that the reconstructed data remain close to the true transitionthis http URL, we prove that the prior-knowledge-based regularization term narrows the solution space from a broad set containing spurious solutions to a constrained subset that satisfies both gradient matching and true transitionthis http URLexperiments on control tasks and autonomous driving tasks demonstrate that RGIA can effectively constrain reconstructed data transition distributions and thus successfully reconstruct local private data.