Improving White-box Robustness of Pre-processing Defenses via Joint Adversarial Training
This addresses a critical problem for practitioners using scalable pre-processing defenses against adversarial attacks, though it is incremental as it builds on existing adversarial training methods.
The paper tackles the robustness degradation effect in pre-processing defenses for deep neural networks, where defenses reduce adversarial robustness in white-box settings, and proposes Joint Adversarial Training based Pre-processing (JATP) to mitigate this effect, showing effectiveness across different target models compared to previous state-of-the-art approaches.
Deep neural networks (DNNs) are vulnerable to adversarial noise. A range of adversarial defense techniques have been proposed to mitigate the interference of adversarial noise, among which the input pre-processing methods are scalable and show great potential to safeguard DNNs. However, pre-processing methods may suffer from the robustness degradation effect, in which the defense reduces rather than improving the adversarial robustness of a target model in a white-box setting. A potential cause of this negative effect is that adversarial training examples are static and independent to the pre-processing model. To solve this problem, we investigate the influence of full adversarial examples which are crafted against the full model, and find they indeed have a positive impact on the robustness of defenses. Furthermore, we find that simply changing the adversarial training examples in pre-processing methods does not completely alleviate the robustness degradation effect. This is due to the adversarial risk of the pre-processed model being neglected, which is another cause of the robustness degradation effect. Motivated by above analyses, we propose a method called Joint Adversarial Training based Pre-processing (JATP) defense. Specifically, we formulate a feature similarity based adversarial risk for the pre-processing model by using full adversarial examples found in a feature space. Unlike standard adversarial training, we only update the pre-processing model, which prompts us to introduce a pixel-wise loss to improve its cross-model transferability. We then conduct a joint adversarial training on the pre-processing model to minimize this overall risk. Empirical results show that our method could effectively mitigate the robustness degradation effect across different target models in comparison to previous state-of-the-art approaches.