🎉 Our paper "Low-Cost Privacy-Aware Decentralized Learning" has been accepted for publication in the Proceedings on Privacy Enhancing Technologies 2025.

Our algorithm uses correlated noise to provide robust privacy protection in decentralized learning while requiring only one communication round per iteration.

Experimental results show up to 35% reduction in membership-inference attacks and 59% higher accuracy compared to existing privacy-preserving methods. Link to the paper.

This would be our second paper at PETS 2025. Link to Shatter.