Summary of Poisoning with a Pill: Circumventing Detection in Federated Learning, by Hanxi Guo et al.
Poisoning with A Pill: Circumventing Detection in Federated Learning
by Hanxi Guo, Hao Wang, Tao Song, Tianhang Zheng, Yang Hua, Haibing Guan, Xiangyu Zhang
First submitted to arxiv on: 22 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); Distributed, Parallel, and Cluster Computing (cs.DC)
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 The proposed approach enhances the effectiveness and stealthiness of existing federated learning (FL) poisoning attacks against detection. The method employs a three-stage methodology that constructs, generates, and injects poison into a pill (a tiny subnet with a novel structure) during the FL training process. This augmentation approach can bypass popular defenses, increasing error rates by up to 7x on IID data and over 2x on non-IID data in both cross-silo and cross-device FL systems. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated learning is great for protecting people’s private data. But some sneaky attacks can try to ruin the process. To stop these attacks, researchers have developed ways to detect them. However, these defenses aren’t perfect. In fact, they’re quite easy to trick. This paper shows that existing defenses are flawed and proposes a new way to make FL poisoning attacks more powerful and harder to detect. |
Keywords
» Artificial intelligence » Federated learning
