Adversarial Immunization for Certifiable Robustness on Graphs

TL;DR

This paper introduces a novel immunization strategy called AdvImmune that enhances the certifiable robustness of graph neural networks against adversarial attacks by vaccinating a small fraction of node pairs, using a meta-gradient optimization approach.

Contribution

It formulates the graph adversarial immunization problem and proposes an efficient algorithm, AdvImmune, to improve GNN robustness with limited immune budget.

Findings

AdvImmune increases robust node ratio significantly on tested networks.

The method achieves robustness improvements with only 5% immune edge budget.

Experimental results show up to 65% increase in robust nodes.

Abstract

Despite achieving strong performance in semi-supervised node classification task, graph neural networks (GNNs) are vulnerable to adversarial attacks, similar to other deep learning models. Existing researches focus on developing either robust GNN models or attack detection methods against adversarial attacks on graphs. However, little research attention is paid to the potential and practice of immunization to adversarial attacks on graphs. In this paper, we propose and formulate the graph adversarial immunization problem, i.e., vaccinating an affordable fraction of node pairs, connected or unconnected, to improve the certifiable robustness of graph against any admissible adversarial attack. We further propose an effective algorithm, called AdvImmune, which optimizes with meta-gradient in a discrete way to circumvent the computationally expensive combinatorial optimization when solving the adversarial immunization problem. Experiments are conducted on two citation networks and one social network. Experimental results demonstrate that the proposed AdvImmune method remarkably improves the ratio of robust nodes by 12%, 42%, 65%, with an affordable immune budget of only 5% edges.