New NL-SME method enhances gradient inversion attacks in federated learning

Researchers have developed NL-SME, a novel method designed to counter multi-step gradient inversion attacks in federated learning. This approach constructs a learnable nonlinear surrogate trajectory to approximate hidden local states, integrating trajectory-level information with calibrated gradient matching. NL-SME also incorporates an update-reliability-aware strategy to mitigate the impact of unreliable components in perturbed updates. Experiments demonstrate that NL-SME significantly outperforms existing methods in reconstruction quality and accuracy, highlighting potential privacy leakage risks even with multi-step updates in federated learning. AI

IMPACT This research highlights potential privacy vulnerabilities in federated learning, prompting further investigation into robust defense mechanisms.

RANK_REASON The cluster contains an academic paper detailing a new method for gradient inversion attacks in federated learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• FedAvg
• Federated Learning
• Li Xia
• NL-SME

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →