Summary of Data Poisoning and Leakage Analysis in Federated Learning, by Wenqi Wei et al.
Data Poisoning and Leakage Analysis in Federated Learning
by Wenqi Wei, Tiansheng Huang, Zachary Yahn, Anoop Singhal, Margaret Loper, Ling Liu
First submitted to arxiv on: 19 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This research chapter delves into the challenges posed by data poisoning and leakage risks in deploying federated learning at scale. The authors highlight two primary threats: training data privacy intrusion and training data poisoning. They investigate how training data can be leaked during federated training and propose a defense strategy involving controlled randomized noise added to gradient updates. The study also reviews various data poisoning attacks, categorizes them, and analyzes their mitigation techniques. Furthermore, the authors demonstrate the potential of dynamic model perturbation in ensuring privacy protection, poisoning resilience, and maintaining model performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated learning is a way for different devices or machines to work together using shared data without sharing it directly. But this approach has some big problems! Data can be leaked or poisoned, which means the results won’t be accurate or trustworthy. The authors of this chapter looked at two main issues: how data might be leaked during training and how data can be intentionally ruined to make the model perform badly. They found that adding a little bit of noise to the updates can help keep the data private. They also examined different types of attacks and showed how to stop them. The chapter concludes by highlighting some other risks, like biased data and misinformation. |
Keywords
» Artificial intelligence » Federated learning
