Defending adversarial attacks in Graph Neural Networks via tensor enhancement

Jianfu Zhang, Yan Hong, Dawei Cheng, Liqing Zhang, Qibin Zhao

Abstract: Highlights•Enhances GNN robustness by low-rank representation of diverse robustness features.•Balances graph structural integrity with improved robustness.•Addresses inter-edge dependencies ignored by current defenses, boosting robustness.•Outperforms leading defense mechanisms against various adversarial attacks.