Summary of Geminio: Language-guided Gradient Inversion Attacks in Federated Learning, by Junjie Shan et al.
Geminio: Language-Guided Gradient Inversion Attacks in Federated Learning
by Junjie Shan, Ziqi Zhao, Jialin Lu, Rui Zhang, Siu Ming Yiu, Ka-Ho Chow
First submitted to arxiv on: 22 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
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 introduces Geminio, a novel approach to transform gradient inversion attacks (GIAs) within federated learning (FL) into semantically meaningful, targeted attacks. The authors leverage foundation models that bridge vision and language to guide the optimization of a malicious global model that prioritizes reconstruction of high-value data samples based on natural language queries. This allows attackers to describe the types of data they consider valuable, and Geminio will focus on reconstructing those samples. The paper demonstrates the effectiveness of Geminio in pinpointing and reconstructing targeted samples across complex datasets under FL and large batch sizes, showing resilience against existing defenses. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research looks at how powerful computer models can be used to steal private data from people who share their information online. The authors create a new way for hackers to find specific pieces of data that are most valuable to them, by using a special kind of model that understands both pictures and words. This allows hackers to describe what they’re looking for in simple language, and the model will help them get it. The authors test their method on large datasets and show that it’s very good at finding the right information. |
Keywords
* Artificial intelligence * Federated learning * Optimization
