IEEE S&P 2025 Paper Introduction

Our paper “Not All Edges are Equally Robust: Evaluating the Robustness of Ranking-Based Federated Learning” has been accepted to the IEEE S&P 2025 — one of the top-tier conferences in cybersecurity and privacy research! 🎉🎉

In this work, we investigate the security limitations of ranking-based Federated Learning (FL) and uncover that it is not inherently robust, with certain edges being significantly more vulnerable to poisoning attacks. To expose and exploit these vulnerabilities, we propose Vulnerable Edge Manipulation (VEM) — a novel local model poisoning strategy that selectively targets the most susceptible components in the FRL. VEM outperforms previous methods by 3.7×, achieving over 53% impact on standard benchmarks. 🔥

📄 Paper

💻 Code

🌐 Project Website

🔗 Check out the LinkedIn post