Game-theoretic Understanding of Adversarially Learned Features.
Year: 2021
Type: preprint
Abstract: This paper aims to understand adversarial attacks and defense from a new perspecitve, i.e., the signal-processing behavior of DNNs. We novelly define the multi-order interaction in game theory, which satisfies six properties. With the multi-order interaction, we discover that adversarial attacks mainly affect high-order interactions to fool the DNN. Furthermore, we find that the robustness of adversarially trained DNNs comes from category-specific low-order interactions. Our findings provide more insights into and make a revision of previous understanding for the shape bias of adversarially learned features. Besides, the multi-order interaction can also explain the recoverability of adversarial examples.
Source: arXiv (Cornell University)
Authors: Jie Ren, Die Zhang, Yisen Wang, Lu Chen, Zhanpeng Zhou, Xu Cheng, Xin Wang, Yi-Ting Chen, Jie Shi, Quanshi Zhang
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3138813434