Guided Adversarial Attack for Evaluating and Enhancing Adversarial Defenses
This work addresses the problem of evaluating and enhancing adversarial defenses for machine learning models by proposing more effective attack and training methods.
This paper introduces Guided Adversarial Margin Attack (GAMA), an adversarial attack that uses a relaxation term and function mapping of clean images to guide adversary generation, resulting in stronger attacks. They also propose Guided Adversarial Training (GAT), which leverages this relaxation term for both attack generation and training, achieving state-of-the-art performance among single-step defenses.
Advances in the development of adversarial attacks have been fundamental to the progress of adversarial defense research. Efficient and effective attacks are crucial for reliable evaluation of defenses, and also for developing robust models. Adversarial attacks are often generated by maximizing standard losses such as the cross-entropy loss or maximum-margin loss within a constraint set using Projected Gradient Descent (PGD). In this work, we introduce a relaxation term to the standard loss, that finds more suitable gradient-directions, increases attack efficacy and leads to more efficient adversarial training. We propose Guided Adversarial Margin Attack (GAMA), which utilizes function mapping of the clean image to guide the generation of adversaries, thereby resulting in stronger attacks. We evaluate our attack against multiple defenses and show improved performance when compared to existing attacks. Further, we propose Guided Adversarial Training (GAT), which achieves state-of-the-art performance amongst single-step defenses by utilizing the proposed relaxation term for both attack generation and training.