Attacking the Madry Defense Model with $L_1$-based Adversarial Examples
This work addresses the robustness of adversarially trained models in computer vision, but it is incremental as it builds on existing attacks and focuses on a specific competition setup.
The paper tackled the problem of attacking the Madry Defense Model by generating adversarial examples with minimal visual distortion using the elastic-net attack (EAD), which achieved transferability despite high average L∞ distortion, questioning the reliance on L∞ as a sole measure.
The Madry Lab recently hosted a competition designed to test the robustness of their adversarially trained MNIST model. Attacks were constrained to perturb each pixel of the input image by a scaled maximal $L_\infty$ distortion $ε$ = 0.3. This discourages the use of attacks which are not optimized on the $L_\infty$ distortion metric. Our experimental results demonstrate that by relaxing the $L_\infty$ constraint of the competition, the elastic-net attack to deep neural networks (EAD) can generate transferable adversarial examples which, despite their high average $L_\infty$ distortion, have minimal visual distortion. These results call into question the use of $L_\infty$ as a sole measure for visual distortion, and further demonstrate the power of EAD at generating robust adversarial examples.