Academic Journal

method for machine learning backdoor attacks based on defense method based on diffusion models

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: method for machine learning backdoor attacks based on defense method based on diffusion models
Συγγραφείς: WANG Sunping, ZHANG Aiqing, YE Xinrong, WANG Yong
Πηγή: 网络与信息安全学报, Vol 11, Pp 137-148 (2025)
Στοιχεία εκδότη: POSTS&TELECOM PRESS Co., LTD, 2025.
Έτος έκδοσης: 2025
Συλλογή: LCC:Electronic computers. Computer science
Θεματικοί όροι: machine learning, backdoor attack, trigger, diffusion model, Electronic computers. Computer science, QA75.5-76.95
Περιγραφή: Backdoor attacks were recognized as one of the primary security threats faced by machine learning models during the training phase. Although significant progress had been achieved in existing defense methods against backdoor attacks, these approaches were often found to result in a substantial decline in model accuracy on clean test sets. To address this issue, a method named defending against backdoor attacks with diffusion model (DBADM) was proposed. The core idea of this method was to preprocess poisoned samples containing backdoor triggers using a diffusion model before model training. By altering the hidden trigger features in the samples, backdoor attacks were effectively mitigated. Systematic offensive and defensive comparison experiments were conducted on four benchmark datasets: MNIST, CIFAR-10, Tiny ImageNet, and LFW. The experimental results demonstrate that the DBADM method not only successfully defends against various backdoor attacks but also maintains the model’s high accuracy performance on the clean dataset.
Τύπος εγγράφου: article
Περιγραφή αρχείου: electronic resource
Γλώσσα: English
Chinese
ISSN: 2096-109X
Relation: http://www.cjnis.com.cn/thesisDetails#10.11959/j.issn.2096-109x.2025059; https://doaj.org/toc/2096-109X
DOI: 10.11959/j.issn.2096-109x.2025059
Σύνδεσμος πρόσβασης: https://doaj.org/article/652580024f434cb3bee22b51cc60144a
Αριθμός Καταχώρησης: edsdoj.652580024f434cb3bee22b51cc60144a
Βάση Δεδομένων: Directory of Open Access Journals