Authors: Yang Zhao, Jiaxi Yang, Yiling Tao, Lixu Wang, Xiaoxiao Li, Dusit Niyato

Published on: October 30, 2023

Impact Score: 8.3

Arxiv code: Arxiv:2310.19218

Summary

• What is new: This survey paper introduces Federated Unlearning (FU) as a new field targeting the ‘right to be forgotten’ in decentralized Federated Learning systems, providing a structured framework to evaluate various FU methods.
• Why this is important: The main challenge is implementing the right to be forgotten in privacy-preserving Federated Learning due to its decentralized nature, requiring a balance between privacy, security, utility, and efficiency.
• What the research proposes: The paper proposes analyzing and unifying existing Federated Unlearning methods into an experimental framework to assess their effectiveness in balancing the mentioned trade-offs.
• Results: A comprehensive analysis of FU methods and their evaluation metrics, along with the creation of a continually updated open-source repository for related materials.

Technical Details

Technological frameworks used: Federated Learning, Federated Unlearning

Models used: Varied FU methods

Data used: nan

Potential Impact

Companies engaged in decentralized data processing, Privacy-oriented tech businesses, Legal and compliance-oriented markets

Want to implement this idea in a business?

We have generated a startup concept here: UniForget.