ASK: Adversarial Soft k-Nearest Neighbor Attack and Defense
This work addresses the vulnerability of kNN-based classification models, which are widely used for their simplicity and interpretability, by introducing both effective attacks and defenses, representing a significant but incremental advance in adversarial machine learning for this specific domain.
The paper tackles the underdeveloped robustness of kNN-based deep learning models by proposing an Adversarial Soft kNN (ASK) loss, which leads to a novel attack method (ASK-Atk) that improves attack success rates by ≥13% on CIFAR-10 and ImageNet, and a defense method (ASK-Def) that enhances robustness by ≥6.9% and ≥3.5% over conventional adversarial training.
K-Nearest Neighbor (kNN)-based deep learning methods have been applied to many applications due to their simplicity and geometric interpretability. However, the robustness of kNN-based classification models has not been thoroughly explored and kNN attack strategies are underdeveloped. In this paper, we propose an Adversarial Soft kNN (ASK) loss to both design more effective kNN attack strategies and to develop better defenses against them. Our ASK loss approach has two advantages. First, ASK loss can better approximate the kNN's probability of classification error than objectives proposed in previous works. Second, the ASK loss is interpretable: it preserves the mutual information between the perturbed input and the in-class-reference data. We use the ASK loss to generate a novel attack method called the ASK-Attack (ASK-Atk), which shows superior attack efficiency and accuracy degradation relative to previous kNN attacks. Based on the ASK-Atk, we then derive an ASK-\underline{Def}ense (ASK-Def) method that optimizes the worst-case training loss induced by ASK-Atk. Experiments on CIFAR-10 (ImageNet) show that (i) ASK-Atk achieves $\geq 13\%$ ($\geq 13\%$) improvement in attack success rate over previous kNN attacks, and (ii) ASK-Def outperforms the conventional adversarial training method by $\geq 6.9\%$ ($\geq 3.5\%$) in terms of robustness improvement.