Dynamics-aware Adversarial Attack of Adaptive Neural Networks
This addresses a critical vulnerability in adaptive neural networks, which are used for computational efficiency in applications like image and point cloud processing, by developing a novel attack method that overcomes the lagged gradient issue, representing a domain-specific advancement.
The paper tackles the problem of adversarial attacks on adaptive neural networks, which change architecture during execution, by proposing a Leaded Gradient Method (LGM) that accounts for dynamic changes to improve attack effectiveness, achieving impressive performance compared to dynamics-unaware methods in experiments on 2D images and 3D point clouds.
In this paper, we investigate the dynamics-aware adversarial attack problem of adaptive neural networks. Most existing adversarial attack algorithms are designed under a basic assumption -- the network architecture is fixed throughout the attack process. However, this assumption does not hold for many recently proposed adaptive neural networks, which adaptively deactivate unnecessary execution units based on inputs to improve computational efficiency. It results in a serious issue of lagged gradient, making the learned attack at the current step ineffective due to the architecture change afterward. To address this issue, we propose a Leaded Gradient Method (LGM) and show the significant effects of the lagged gradient. More specifically, we reformulate the gradients to be aware of the potential dynamic changes of network architectures, so that the learned attack better "leads" the next step than the dynamics-unaware methods when network architecture changes dynamically. Extensive experiments on representative types of adaptive neural networks for both 2D images and 3D point clouds show that our LGM achieves impressive adversarial attack performance compared with the dynamic-unaware attack methods. Code is available at https://github.com/antao97/LGM.