Provably Robust Explainable Graph Neural Networks against Graph Perturbation Attacks

2025-02-06

AI Summary

Existing eXplainable Graph Neural Networks (XGNNs) suffer from poor robustness—explanations exhibit significant instability under minor perturbations to graph structure. Method: We propose XGNNCert, the first provably robust explanation framework for GNNs. It formally defines provable robustness of XGNNs against bounded edge perturbations, integrates certification theory with sensitivity analysis to enforce explanation stability, and remains compatible with mainstream explainers (e.g., GNNExplainer, PGM-Explainer). Contribution/Results: XGNNCert guarantees explanation stability under worst-case structural perturbations without degrading the original GNN’s predictive accuracy. Extensive experiments on benchmark datasets demonstrate that XGNNCert improves average perturbation tolerance by 3.2× over baselines, substantially enhancing explanation reliability. This work establishes the first theoretical foundation and practical solution for robust, certifiable GNN interpretability—critical for safety-critical applications.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Explaining Graph Neural Network (XGNN) has gained growing attention to facilitate the trust of using GNNs, which is the mainstream method to learn graph data. Despite their growing attention, Existing XGNNs focus on improving the explanation performance, and its robustness under attacks is largely unexplored. We noticed that an adversary can slightly perturb the graph structure such that the explanation result of XGNNs is largely changed. Such vulnerability of XGNNs could cause serious issues particularly in safety/security-critical applications. In this paper, we take the first step to study the robustness of XGNN against graph perturbation attacks, and propose XGNNCert, the first provably robust XGNN. Particularly, our XGNNCert can provably ensure the explanation result for a graph under the worst-case graph perturbation attack is close to that without the attack, while not affecting the GNN prediction, when the number of perturbed edges is bounded. Evaluation results on multiple graph datasets and GNN explainers show the effectiveness of XGNNCert.

Problem

• Ensures robustness of XGNN against attacks
• Proposes XGNNCert for provable robustness
• Maintains explanation accuracy under perturbations

Innovation

• Provably robust XGNN
• Ensures explanation consistency
• Resists graph perturbation attacks