Toward Adversarial Robustness via Semi-supervised Robust Training
This work addresses adversarial robustness for deep learning models, offering a semi-supervised approach that improves defense against various perturbations, though it appears incremental as it builds on adversarial training concepts.
The paper tackles the problem of adversarial examples threatening deep neural networks by proposing a novel defense method called robust training (RT) and its semi-supervised extension (SRT), which jointly minimizes standard and robust risks to enhance accuracy and robustness; experiments show SRT outperforms state-of-the-art methods on benchmark datasets for defending against pixel-wise and spatial perturbations.
Adversarial examples have been shown to be the severe threat to deep neural networks (DNNs). One of the most effective adversarial defense methods is adversarial training (AT) through minimizing the adversarial risk $R_{adv}$, which encourages both the benign example $x$ and its adversarially perturbed neighborhoods within the $\ell_{p}$-ball to be predicted as the ground-truth label. In this work, we propose a novel defense method, the robust training (RT), by jointly minimizing two separated risks ($R_{stand}$ and $R_{rob}$), which is with respect to the benign example and its neighborhoods respectively. The motivation is to explicitly and jointly enhance the accuracy and the adversarial robustness. We prove that $R_{adv}$ is upper-bounded by $R_{stand} + R_{rob}$, which implies that RT has similar effect as AT. Intuitively, minimizing the standard risk enforces the benign example to be correctly predicted, and the robust risk minimization encourages the predictions of the neighbor examples to be consistent with the prediction of the benign example. Besides, since $R_{rob}$ is independent of the ground-truth label, RT is naturally extended to the semi-supervised mode ($i.e.$, SRT), to further enhance the adversarial robustness. Moreover, we extend the $\ell_{p}$-bounded neighborhood to a general case, which covers different types of perturbations, such as the pixel-wise ($i.e.$, $x + δ$) or the spatial perturbation ($i.e.$, $ AX + b$). Extensive experiments on benchmark datasets not only verify the superiority of the proposed SRT method to state-of-the-art methods for defensing pixel-wise or spatial perturbations separately, but also demonstrate its robustness to both perturbations simultaneously. The code for reproducing main results is available at \url{https://github.com/THUYimingLi/Semi-supervised_Robust_Training}.