Enhancing Output Diversity Improves Conjugate Gradient-based Adversarial Attacks
This work addresses the vulnerability of deep neural networks to adversarial examples, offering an incremental improvement in attack methods for security testing.
The paper tackled the problem of improving adversarial attack performance by enhancing output diversity, proposing ReACG which automatically modifies search direction and step size to increase distance between consecutive search points, resulting in higher attack performance than ACG, particularly effective for ImageNet models with multiple classes.
Deep neural networks are vulnerable to adversarial examples, and adversarial attacks that generate adversarial examples have been studied in this context. Existing studies imply that increasing the diversity of model outputs contributes to improving the attack performance. This study focuses on the Auto Conjugate Gradient (ACG) attack, which is inspired by the conjugate gradient method and has a high diversification performance. We hypothesized that increasing the distance between two consecutive search points would enhance the output diversity. To test our hypothesis, we propose Rescaling-ACG (ReACG), which automatically modifies the two components that significantly affect the distance between two consecutive search points, including the search direction and step size. ReACG showed higher attack performance than that of ACG, and is particularly effective for ImageNet models with several classification classes. Experimental results show that the distance between two consecutive search points enhances the output diversity and may help develop new potent attacks. The code is available at \url{https://github.com/yamamura-k/ReACG}