Boosting Graph Robustness Against Backdoor Attacks: An Over-Similarity Perspective

TL;DR

This paper introduces SimGuard, a novel defense method against graph backdoor attacks that leverages over-similarity analysis and contrastive learning to effectively detect and mitigate triggers in GNNs.

Contribution

The paper presents a new similarity-based detection approach and contrastive learning framework to improve robustness of GNNs against backdoor attacks.

Findings

SimGuard effectively detects backdoor triggers in GNNs.

The method preserves performance on clean nodes.

Extensive experiments validate the approach's effectiveness.

Abstract

Graph Neural Networks (GNNs) have achieved notable success in tasks such as social and transportation networks. However, recent studies have highlighted the vulnerability of GNNs to backdoor attacks, raising significant concerns about their reliability in real-world applications. Despite initial efforts to defend against specific graph backdoor attacks, existing defense methods face two main challenges: either the inability to establish a clear distinction between triggers and clean nodes, resulting in the removal of many clean nodes, or the failure to eliminate the impact of triggers, making it challenging to restore the target nodes to their pre-attack state. Through empirical analysis of various existing graph backdoor attacks, we observe that the triggers generated by these methods exhibit over-similarity in both features and structure. Based on this observation, we propose a novel graph backdoor defense method SimGuard. We first utilizes a similarity-based metric to detect triggers and then employs contrastive learning to train a backdoor detector that generates embeddings capable of separating triggers from clean nodes, thereby improving detection efficiency. Extensive experiments conducted on real-world datasets demonstrate that our proposed method effectively defends against various graph backdoor attacks while preserving performance on clean nodes. The code will be released upon acceptance.