On the Effectiveness of Adversarial Training against Backdoor Attacks
This addresses the problem of securing DNNs against backdoor attacks for practitioners relying on internet-collected data, offering an incremental improvement in defense strategies.
The paper investigates whether adversarial training can defend against backdoor attacks in DNNs, finding that the threat model matters, with spatial adversarial examples providing notable robustness against patch-based attacks, and proposes a hybrid strategy for satisfactory robustness across different attacks.
DNNs' demand for massive data forces practitioners to collect data from the Internet without careful check due to the unacceptable cost, which brings potential risks of backdoor attacks. A backdoored model always predicts a target class in the presence of a predefined trigger pattern, which can be easily realized via poisoning a small amount of data. In general, adversarial training is believed to defend against backdoor attacks since it helps models to keep their prediction unchanged even if we perturb the input image (as long as within a feasible range). Unfortunately, few previous studies succeed in doing so. To explore whether adversarial training could defend against backdoor attacks or not, we conduct extensive experiments across different threat models and perturbation budgets, and find the threat model in adversarial training matters. For instance, adversarial training with spatial adversarial examples provides notable robustness against commonly-used patch-based backdoor attacks. We further propose a hybrid strategy which provides satisfactory robustness across different backdoor attacks.