Summary of Resilience in Online Federated Learning: Mitigating Model-poisoning Attacks Via Partial Sharing, by Ehsan Lari et al.
Resilience in Online Federated Learning: Mitigating Model-Poisoning Attacks via Partial Sharing
by Ehsan Lari, Reza Arablouei, Vinay Chakravarthi Gogineni, Stefan Werner
First submitted to arxiv on: 19 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); Distributed, Parallel, and Cluster Computing (cs.DC); Signal Processing (eess.SP)
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 paper investigates the resilience of Federated Learning (FL) against model-poisoning attacks, a type of attack where malicious clients tamper with their local models to manipulate the global model. The partial-sharing online FL (PSO-Fed) algorithm is studied for its ability to withstand these attacks. PSO-Fed reduces communication overhead by allowing clients to share only a fraction of their model updates with the server, which surprisingly enhances its robustness to model-poisoning attacks. Theoretical analysis shows that PSO-Fed maintains convergence even under Byzantine attacks, where malicious clients inject noise into their updates. A formula is derived for PSO-Fed’s mean square error, considering factors like stepsize, attack probability, and the number of malicious clients. An optimal stepsize is found that maximizes PSO-Fed’s resistance to these attacks. Extensive numerical experiments confirm the theoretical findings, showcasing PSO-Fed’s superior performance against model-poisoning attacks compared to other leading FL algorithms. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper looks at how a type of machine learning called Federated Learning (FL) can be attacked by bad actors. FL allows many devices to work together to learn from each other without sharing their data. But, if someone wants to cheat the system, they can manipulate the information shared between devices. The researchers studied an algorithm that reduces the amount of information shared and found it makes it harder for attackers to succeed. They also showed that this algorithm works even when some attackers are trying to make it fail. This is important because it helps keep our data safe online. |
Keywords
* Artificial intelligence * Federated learning * Machine learning * Probability
