GameRLFed: Robust Federated Learning Against Evolving Backdoor Attacks via Game Theory and Deep Reinforcement Learning

Summary

GameRLFed is a defense framework against backdoor attacks in federated learning, where compromised clients send malicious updates that cause the global model to misbehave on trigger inputs. It models the problem as a repeated game with incomplete information, and the server uses reinforcement learning to adjust aggregation weights without auxiliary data. The authors report that it outperforms six baseline methods in robustness, adaptability and computational efficiency across multiple datasets and attack types.

Topics

• Adversarial machine learning

Related items

• MediumTrained to betray: How post-training supply chain attacks hijack coding agentsSimilar attack · Microsoft Security Blog
• MediumRewriting less than 10 neurons in an LLM can turn it into a sleeper backdoorSimilar attack · CSO Online
• InfoLazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI AgentsSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
• InfoHijacking a Trusted LLM in Real-World Quantized Deployments Only Requires a Disguised AdapterSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
• InfoPoison-then-Hide: Finetuning-Activated Backdoor Attack on Pretrained Vision EncodersSimilar attack · Arxiv (cs.RO + cs.CV security)