BadSFL: Backdoor Attack against Scaffold Federated Learning

Federated learning (FL) enables the training of deep learning models on distributed clients to preserve data privacy. However, this learning paradigm is vulnerable to backdoor attacks, where malicious clients can upload poisoned local models to embed backdoors into the global model, leading to attacker-desired predictions. Existing backdoor attacks mainly focus on FL with independently and identically distributed (IID) scenarios, while real-world FL training data are typically non-IID. Current strategies for non-IID backdoor attacks suffer from limitations in maintaining effectiveness and durability. To address these challenges, we propose a novel backdoor attack method, BadSFL, specifically designed for the FL framework using the scaffold aggregation algorithm in non-IID settings. BadSFL leverages a Generative Adversarial Network (GAN) based on the global model to complement the training set, achieving high accuracy on both backdoor and benign samples. It utilizes a specific feature as the backdoor trigger to ensure stealthiness, and exploits the Scaffold's control variate to predict the global model's convergence direction, ensuring the backdoor's persistence. Extensive experiments on three benchmark datasets demonstrate the high effectiveness, stealthiness, and durability of BadSFL. Notably, our attack remains effective over 60 rounds in the global model and up to 3 times longer than existing baseline attacks after stopping the injection of malicious updates.

Clean-Annotation Backdoor Attack against Lane Detection Systems in the Wild

We present the first backdoor attack against the lane detection systems in the physical world. Modern autonomous vehicles adopt various deep learning methods to train lane detection models, making it challenging to devise a universal backdoor attack technique. In our solution, (1) we propose a novel semantic trigger design, which leverages the traffic cones with specific poses and locations to activate the backdoor. Such trigger can be easily realized under the physical setting, and looks very natural not to be detected. (2) We introduce a new clean-annotation approach to generate poisoned samples. These samples have correct annotations but are still capable of embedding the backdoor to the model. Comprehensive evaluations on public datasets and physical autonomous vehicles demonstrate that our backdoor attack is effective, stealthy and robust.