Summary of Vflip: a Backdoor Defense For Vertical Federated Learning Via Identification and Purification, by Yungi Cho et al.
VFLIP: A Backdoor Defense for Vertical Federated Learning via Identification and Purification
by Yungi Cho, Woorim Han, Miseon Yu, Younghan Lee, Ho Bae, Yunheung Paek
First submitted to arxiv on: 28 Aug 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 paper presents Vertical Federated Learning (VFL) backdoor defense, specifically designed for VFL. The authors identify a significant vulnerability in VFL to backdoor attacks that target its distinct characteristics. These attacks can neutralize existing defense mechanisms designed for Horizontal Federated Learning (HFL) and deep neural networks. To address this issue, the authors propose VFLIP, a novel backdoor defense approach that operates at the inference stage. VFLIP uses participant-wise anomaly detection to identify malicious embeddings and then purifies them by removing identified malicious ones and reconstructing all embeddings based on remaining ones. The paper demonstrates VFLIP’s effectiveness in mitigating backdoor attacks on various datasets, including CIFAR10, CINIC10, Imagenette, NUS-WIDE, and BankMarketing. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about keeping data safe when it’s shared across different groups. Right now, there’s a big problem with how this sharing works called backdoor attacks. These attacks can ruin the whole system and make it hard to tell what’s real and what’s not. The authors of this paper came up with a new way to stop these attacks, which they call VFLIP. It works by finding the bad parts of the data and getting rid of them. They tested it on different types of data and showed that it really works! |
Keywords
» Artificial intelligence » Anomaly detection » Federated learning » Inference
