New gradient inversion attack reveals significant privacy risks in federated learning

Researchers have developed a new method for gradient inversion attacks in federated learning, inspired by LT codes. This technique allows for exact recovery of training data and labels from a single round of FedSGD, significantly outperforming previous single-round attacks. Even passive attackers can recover a high percentage of batches from benchmarks like ImageNet, suggesting that the privacy risks in federated learning have been underestimated. AI

IMPACT This research highlights significant privacy vulnerabilities in federated learning, potentially impacting how data is shared and secured in distributed AI systems.

RANK_REASON The cluster contains a research 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.AI →

• alphaXiv
• arXiv
• DagsHub
• federated learning
• FedSGD
• Hugging Face
• IArxiv
• ImageNet
• LT Codes and the Minimum Spanning Tree Based Distributed Storage in Wireless Sensor Networks
• Mohsen Alambardar Meybodi

• paper
• safety

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