We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CR

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Cryptography and Security

Title: Black-box Detection of Backdoor Attacks with Limited Information and Data

Abstract: Although deep neural networks (DNNs) have made rapid progress in recent years, they are vulnerable in adversarial environments. A malicious backdoor could be embedded in a model by poisoning the training dataset, whose intention is to make the infected model give wrong predictions during inference when the specific trigger appears. To mitigate the potential threats of backdoor attacks, various backdoor detection and defense methods have been proposed. However, the existing techniques usually require the poisoned training data or access to the white-box model, which is commonly unavailable in practice. In this paper, we propose a black-box backdoor detection (B3D) method to identify backdoor attacks with only query access to the model. We introduce a gradient-free optimization algorithm to reverse-engineer the potential trigger for each class, which helps to reveal the existence of backdoor attacks. In addition to backdoor detection, we also propose a simple strategy for reliable predictions using the identified backdoored models. Extensive experiments on hundreds of DNN models trained on several datasets corroborate the effectiveness of our method under the black-box setting against various backdoor attacks.
Subjects: Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Machine Learning (stat.ML)
Cite as: arXiv:2103.13127 [cs.CR]
  (or arXiv:2103.13127v1 [cs.CR] for this version)

Submission history

From: Yinpeng Dong [view email]
[v1] Wed, 24 Mar 2021 12:06:40 GMT (802kb,D)

Link back to: arXiv, form interface, contact.