Model Poisoning Attack Detection in Federated Learning

Federated learning is a distributed machine learning technique that aggregates every client model on the server side. There can be various types of attacks to destroy the robustness of this learning system. A recent study* introduces a low-cost approach for the server to detect these malicious models by coordinate-based statistical comparison. In this project, we will extend this method for detecting model poisoning attacks both on the clients and on the server.

*: C. Çağlayan and A. Yurdakul, “A Clustering-Based Scoring Mechanism for Malicious Model Detection in Federated Learning,” 25th Euromicro Conference on Digital System Design (DSD’22), August 31 - September 2, 2022, Gran Canaria, Spain.