Adversarial Robustness without Adversarial Training: A Teacher-Guided Curriculum Learning Approach

Anindya Sarkar, Anirban Sarkar, Sowrya Gali, Vineeth N. Balasubramanian

Keywords: Adversarial Robustness, Inner Maximization, Outer Minimization, Attribution Map, Curriculum Learning

TL;DR: We propose a non-iterative robust training technique which takes 10-20% time over existing adversarial training (AT) based models and outperforms strong baselines for both adversarial as well as natural accuracies.

Abstract: Current SOTA adversarially robust models are mostly based on adversarial training (AT) and differ only by some regularizers either at inner maximization or outer minimization steps. Being repetitive in nature during the inner maximization step, they take a huge time to train. We propose a non-iterative method that enforces the following ideas during training. Attribution maps are more aligned to the actual object in the image for adversarially robust models compared to naturally trained models. Also, the allowed set of pixels to perturb an image (that changes model decision) should be restricted to the object pixels only, which reduces the attack strength by limiting the attack space. Our method achieves significant performance gains with a little extra effort (10-20%) over existing AT models and outperforms all other methods in terms of adversarial as well as natural accuracy. We have performed extensive experimentation with CIFAR-10, CIFAR-100, and TinyImageNet datasets and reported results against many popular strong adversarial attacks to prove the effectiveness of our method.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

Code: https://github.com/sowgali/Get-Fooled-for-the-Right-Reason