<i>Model X-ray</i> : Detecting Backdoored Models via Decision Boundary
Backdoor attacks pose a significant security vulnerability for deep neural networks (DNNs), enabling them to operate normally on clean inputs but manipulate predictions when specific trigger patterns occur. In this paper, we consider a practical post-training scenario backdoor defense, where the defender aims to evaluate whether a trained model has been compromised by backdoor attacks. Currently, post-training backdoor detection approaches often operate under the assumption that the defender has knowledge of the attack information, logit output from the model, and knowledge of the model parameters, limiting their implementation in practical scenarios.
Authors
- Nenghai Yu (ORCID: https://orcid.org/0000-0003-4417-9316)
- Weiming Zhang (ORCID: https://orcid.org/0000-0001-5576-6108)
- Jie Zhang (ORCID: https://orcid.org/0000-0002-4230-1077)
- Yanghao Su
- Tianwei Zhang (ORCID: https://orcid.org/0000-0001-6595-6650)
- Ting Xu (ORCID: https://orcid.org/0000-0002-5316-7213)
Institutions
- Agency for Science, Technology and Research (SG)
- University of Science and Technology of China (CN)
- National University of Singapore (SG)
- Nanyang Technological University (SG)
Publication Details
- Published
- 2024-10-26
- DOI
- https://doi.org/10.1145/3664647.3681075
- Primary Topic
- Adversarial Robustness in Machine Learning
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Research Foundation
- National Research Foundation Singapore
- National Natural Science Foundation of China
- Ministry of Education, India
- Info-communications Media Development Authority