BadVFL: Backdoor Attacks in Vertical Federated Learning
Mohammad Naseri, Yufei Han, Emiliano De Cristofaro
IEEE Symposium on Security and Privacy 2024 · Day 2 · Continental Ballroom 5
This talk, "BadVFL: Backdoor Attacks in Vertical Federated Learning," presented by Mohammad Naseri, Yufei Han, and Emiliano De Cristofaro, delves into a novel class of adversarial attacks targeting **Vertical Federated Learning (VFL)** systems. Federated Learning (FL) has emerged as a crucial privacy-preserving machine learning paradigm, allowing multiple parties to collaboratively train a shared model without directly exchanging their sensitive raw data. While the security implications in Horizontal Federated Learning (HFL) have been extensively studied, VFL, with its distinct architectural and data distribution characteristics, has received comparatively limited attention regarding robustness attacks.
AI review
This presentation on BadVFL delivers a critical, novel backdoor attack against Vertical Federated Learning, a paradigm often assumed robust due to its architectural split. The researchers demonstrate a sophisticated two-stage, clean-label technique that bypasses VFL's unique constraints, proving traditional HFL defenses are inadequate. This work is essential for anyone deploying or researching VFL, exposing a significant vulnerability that demands immediate attention.