Adverseness vs. Equilibrium: Exploring Graph Adversarial Resilience through Dynamic Equilibrium
This addresses the problem of improving graph adversarial resilience for researchers and practitioners in graph analytics, representing a novel theoretical and methodological contribution rather than an incremental improvement.
The paper tackles the problem of identifying an intrinsic adversarial resilience state in graphs by modeling adversarial learning as a dynamic system, proposing a theoretical framework to show its existence, and developing a method to pinpoint this state through equilibrium points. The result is that their approach significantly outperforms state-of-the-art defense methods across five real-world datasets and three attacks.
Adversarial attacks to graph analytics are gaining increased attention. To date, two lines of countermeasures have been proposed to resist various graph adversarial attacks from the perspectives of either graph per se or graph neural networks. Nevertheless, a fundamental question lies in whether there exists an intrinsic adversarial resilience state within a graph regime and how to find out such a critical state if exists. This paper contributes to tackle the above research questions from three unique perspectives: i) we regard the process of adversarial learning on graph as a complex multi-object dynamic system, and model the behavior of adversarial attack; ii) we propose a generalized theoretical framework to show the existence of critical adversarial resilience state; and iii) we develop a condensed one-dimensional function to capture the dynamic variation of graph regime under perturbations, and pinpoint the critical state through solving the equilibrium point of dynamic system. Multi-facet experiments are conducted to show our proposed approach can significantly outperform the state-of-the-art defense methods under five commonly-used real-world datasets and three representative attacks.