Your Privacy My Cloak: Backdoor Attacks on Differentially Private Federated Learning
Researchers have developed a new attack method called RING that exploits differential privacy (DP) in federated learning (FL) to conceal malicious updates. Contrary to prior assumptions, DP can mask the statistical characteristics of backdoor attacks, rendering existing defenses ineffective. RING achieves a 90.3% attack success rate against state-of-the-art defenses, highlighting a significant security vulnerability in DP-FL deployments that comes with substantial utility trade-offs. AI
IMPACT Exposes a fundamental security gap in differentially private federated learning, potentially requiring new defense mechanisms.