GDetox: Purifying Backdoor Encoder in Graph Self-Supervised Learning via Knowledge Distillation

Summary

GDetox is a defense against backdoored graph encoders in graph self-supervised learning (GSSL), the first of its kind per the source. It purifies the encoder through self-supervised distillation without label information, adding adversarial contrastive learning to strengthen the teacher model. Across four node and four graph classification datasets, GDetox reduced the attack success rate to 4% with encoder performance degrading by no more than 2%.

Mitigation

GDetox: a backdoor defense that purifies a backdoored graph encoder via self-supervised distillation, using adversarial contrastive learning to strengthen the teacher model.

Related items

• MediumXbow spent $40K on AI and cracked Google Chrome wide openSimilar attack · CSO Online
• MediumOpenAI Research Model Spent 3rd-Party Test ETH Without Permission to Deploy Smart ContractSimilar attack · The Hacker News
• InfoTADR: Threat-Aligned Data Refinement against label-flipping attacks in AI-powered Intrusion Detection SystemsSimilar attack · Elsevier Security Journals
• MediumOpenAI Models Hacked Data Portals, Raided a Crypto Wallet to Finish TasksSimilar attack · SecurityWeek
• LowMalware evasion made easier with AI — but behavior still gives it awaySimilar attack · CSO Online