Detecting Scene-Plausible Perceptible Backdoors in Trained DNNs Without Access to the Training Set
Abstract
Backdoor data poisoning attacks add mislabeled examples to the training set, with an embedded backdoor pattern, so that the classifier learns to classify to a target class whenever the backdoor pattern is present in a test sample. Here, we address posttraining detection of scene-plausible perceptible backdoors, a type of backdoor attack that can be relatively easily fashioned, particularly against DNN image classifiers. A post-training defender does not have access to the potentially poisoned training set, only to the trained classifier, as well as some unpoisoned examples that need not be training samples. Without the poisoned training set, the only information about a backdoor pattern is encoded in the DNN's trained weights. This detection scenario is of great import considering legacy and proprietary systems, cell phone apps, as well as training outsourcing, where the user of the classifier will not have access to the entire training set. We identify two important properties of scene-plausible perceptible backdoor patterns, spatial invariance and robustness, based on which we propose a novel detector using the maximum achievable misclassification fraction (MAMF) statistic. We detect whether the trained DNN has been backdoor-attacked and infer the source and target classes. Our detector outperforms existing detectors and, coupled with an imperceptible backdoor detector, helps achieve posttraining detection of most evasive backdoors of interest.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The adaptive learning rate method shows reasonable defense performance against data poisoning attacks.
“This method shows reasonable defense performance against data poisoning attacks [214].”
From Data and model poisoning backdoor attacks on wireless federated learning, and the defense mechanisms: A comprehensive survey (2024) · cites this paper as [214]
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · see citing works
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 74,248 · see citing works
Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization
Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., & Batra, D. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.74 · Cited by 24,513 · see citing works
Advances in Neural Information Processing Systems 25
Zhang, Y., Sutton, C. A., Storkey, A., & Zoubin Ghahramani, A. (2012) · Cited by 12,653 · see citing works
Explaining and Harnessing Adversarial Examples
Goodfellow, I. J., Shlens, J., & Szegedy, C. (2014) · arXiv · doi:10.48550/arxiv.1412.6572 · Cited by 10,381 · see citing works
Towards Evaluating the Robustness of Neural Networks
Carlini, N., & Wagner, D. (2017) · 2017 IEEE Symposium on Security and Privacy (SP) · doi:10.1109/sp.2017.49 · Cited by 8,859 · see citing works
Intriguing properties of neural networks
Szegedy, C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I., & Fergus, R. (2013) · arXiv · doi:10.48550/arxiv.1312.6199 · Cited by 8,143 · see citing works
DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks
Moosavi-Dezfooli, S.-M., Fawzi, A., & Frossard, P. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.282 · Cited by 5,317 · see citing works
The Limitations of Deep Learning in Adversarial Settings
Papernot, N., McDaniel, P., Jha, S., Fredrikson, M., Celik, Z. B., & Swami, A. (2016) · 2016 IEEE European Symposium on Security and Privacy (EuroS&P) · doi:10.1109/eurosp.2016.36 · Cited by 3,993 · see citing works
Abe, S. (2001) · Springer eBooks · doi:10.1007/978-1-4471-0285-4 · Cited by 3,740 · see citing works
Practical Black-Box Attacks against Machine Learning
Papernot, N., McDaniel, P., Goodfellow, I., Jha, S., Celik, Z. B., & Swami, A. (2017) · Proceedings of the 2017 ACM on Asia Conference on Computer and Communications Security · doi:10.1145/3052973.3053009 · Cited by 3,495 · see citing works
Universal Adversarial Perturbations
Moosavi-Dezfooli, S.-M., Fawzi, A., Fawzi, O., & Frossard, P. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.17 · Cited by 2,680 · see citing works
The Influence Curve and its Role in Robust Estimation
Hampel, F. R. (1974) · Journal of the American Statistical Association · doi:10.1080/01621459.1974.10482962 · Cited by 2,554 · see citing works
Wild patterns: Ten years after the rise of adversarial machine learning
Biggio, B., & Roli, F. (2018) · Pattern Recognition · doi:10.1016/j.patcog.2018.07.023 · Cited by 1,563 · see citing works
Understanding Black-box Predictions via Influence Functions
Koh, P. W., & Liang, P. (2017) · arXiv · doi:10.48550/arxiv.1703.04730 · Cited by 1,514 · see citing works
Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks
Wang, B., Yao, Y., Shan, S., Li, H., Viswanath, B., Zheng, H.-T., & Zhao, B. Y. (2019) · IEEE Symposium on Security and Privacy (S&P) · doi:10.1109/sp.2019.00031 · Cited by 1,486 · see citing works
Trojaning Attack on Neural Networks
Liu, Y., Ma, S., Aafer, Y., Lee, W.-C., Zhai, J., Wang, W., & Zhang, X. (2018) · Proceedings 2018 Network and Distributed System Security Symposium · doi:10.14722/ndss.2018.23291 · Cited by 1,270 · see citing works
BadNets: Evaluating Backdooring Attacks on Deep Neural Networks
Gu, T., Liu, K., Dolan-Gavitt, B., & Garg, S. (2019) · IEEE Access · doi:10.1109/access.2019.2909068 · Cited by 1,238 · see citing works
Huang, L., Joseph, A. D., Nelson, B., Rubinstein, B. I. P., & Tygar, J. D. (2011) · Proceedings of the 4th ACM workshop on Security and artificial intelligence · doi:10.1145/2046684.2046692 · Cited by 1,194 · see citing works
Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks
Liu, K., Dolan-Gavitt, B., & Garg, S. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-00470-5_13 · Cited by 1,055 · see citing works
Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
Chen, X., Liu, C., Li, B., Lu, K., & Song, D. (2017) · arXiv · doi:10.48550/arxiv.1712.05526 · Cited by 1,039 · see citing works
Gao, Y., Xu, C., Wang, D., Chen, S., Ranasinghe, D. C., & Nepal, . (2019) · Annual Computer Security Applications Conference · doi:10.1145/3359789.3359790 · Cited by 719 · see citing works
EAD: Elastic-Net Attacks to Deep Neural Networks via Adversarial Examples
Chen, P., Sharma, Y., Zhang, H., Yi, J., & Hsieh, C. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.11302 · Cited by 594 · see citing works
Spectral Signatures in Backdoor AttacksPreprint
Tran, B., Li, J. Z., & Mądry, A. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.00636 · Cited by 506 · see citing works
Detecting Backdoor Attacks on Deep Neural Networks by Activation Clustering
Chen, B., Carvalho, W., Baracaldo, N., Ludwig, H., Edwards, B., Lee, T., Molloy, I., & Srivastava, B. (2018) · arXiv · doi:10.48550/arxiv.1811.03728 · Cited by 485 · see citing works
Convolutional Neural Networks Applied to House Numbers Digit ClassificationPreprint
Sermanet, P., Chintala, S., & LeCun, Y. (2012) · arXiv (Cornell University) · doi:10.48550/arxiv.1204.3968 · Cited by 336 · see citing works
Backdoor Embedding in Convolutional Neural Network Models via Invisible Perturbation
Zhong, H., Liao, C., Squicciarini, A., Zhu, S., & Miller, D. J. (2020) · ACM Conference on Data and Application Security and Privacy (CODASPY) · doi:10.1145/3374664.3375751 · Cited by 306 · see citing works
Self-citationMiller, D. J., Xiang, Z. J., & Kesidis, G. (2020) · Proceedings of the IEEE · doi:10.1109/jproc.2020.2970615 · Cited by 238 · see citing works
Self-citationSupport vector machines under adversarial label contamination
Huang, X. X., Biggio, B., Nelson, B. A., Xiao, H., Eckert, C. M., & Roli, F. (2015) · Neurocomputing · doi:10.1016/j.neucom.2014.08.081 · Cited by 220 · see citing works
TABOR: A Highly Accurate Approach to Inspecting and Restoring Trojan Backdoors in AI SystemsPreprint
Guo, W., Wang, L., Xing, X., Du, M., & Song, D. X. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1908.01763 · Cited by 154 · see citing works
Generative Poisoning Attack Method Against Neural NetworksPreprint
Yang, C., Wu, Q., Li, H. H., & Chen, Y. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1703.01340 · Cited by 141 · see citing works
STRIP: A Defence Against Trojan Attacks on Deep Neural NetworksPreprint
Gao, Y., Xu, C., Wang, D., Chen, S., Ranasinghe, D. C., & Nepal, . (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1902.06531 · Cited by 127 · see citing works
Misleading Learners: Co-opting Your Spam Filter
Nelson, B. A., Barreno, M., Chi, F. J., Joseph, A. D., Rubinstein, B. I. P., Saini, U., Sutton, C. A., Tygar, J. D., & Xia, K. (2009) · Springer eBooks · doi:10.1007/978-0-387-88735-7_2 · Cited by 113 · see citing works
SentiNet: Detecting Physical Attacks Against Deep Learning SystemsPreprint
Chou, E., Tramèr, F., Pellegrino, G., & Boneh, D. (2018) · arXiv (Cornell University) · Cited by 76 · see citing works
Xiang, Z. J., Miller, D. J., & Kesidis, G. (2020) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp40776.2020.9054581 · Cited by 33 · see citing works
Self-citationXiang, Z. J., Miller, D. J., & Kesidis, G. (2021) · Computers & Security · doi:10.1016/j.cose.2021.102280 · Cited by 27 · see citing works
Self-citationMiller, D. J., Xiang, Z. J., & Kesidis, G. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1904.06292 · Cited by 22 · see citing works
Self-citationXiang, Z. J., Miller, D. J., & Kesidis, G. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9414562 · Cited by 13 · see citing works
Self-citationXiang, Z. J., Miller, D. J., Wang, H., & Kesidis, G. (2020) · Machine learning for signal processing ... · doi:10.1109/mlsp49062.2020.9231861 · Cited by 9 · see citing works
Self-citationXiang, Z. J., Miller, D. J., Wang, H., & Kesidis, G. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1911.07970 · Cited by 6 · see citing works
Self-citation
Flagged first, then by citations
3 of the 43 references couldn’t be matched to a known paper.
Cited by
Sorted by citations
Wan, Y., Qu, Y., Ni, W., Xiang, Y., Gao, L., & Hossain, E. (2024) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2024.3361451 · Cited by 152 · see citing works
“This method shows reasonable defense performance against data poisoning attacks [214].”
MethodAn Overview of Backdoor Attacks Against Deep Neural Networks and Possible Defences
Guo, W., Tondi, B., & Barni, M. (2022) · IEEE Open Journal of Signal Processing · doi:10.1109/ojsp.2022.3190213 · Cited by 91 · see citing works
Wang, H., Xiang, Z. J., Miller, D. J., & Kesidis, G. (2024) · IEEE Symposium on Security and Privacy (S&P) · doi:10.1109/sp54263.2024.00015 · Cited by 53 · see citing works
Self-citationXiang, Z. J., Miller, D. J., & Kesidis, G. (2021) · Computers & Security · doi:10.1016/j.cose.2021.102280 · Cited by 27 · see citing works
Self-citationShadow backdoor attack: Multi-intensity backdoor attack against federated learning
Ren, Q., Zheng, Y., Chao, Y., Li, Y., & Ma, J. (2024) · Computers & Security · doi:10.1016/j.cose.2024.103740 · Cited by 14 · see citing works
Detecting Backdoor Attacks against Point Cloud Classifiers
Xiang, Z. J., Miller, D. J., Chen, S., Li, X., & Kesidis, G. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9747194 · Cited by 10 · see citing works
Self-citationTraining Set Cleansing of Backdoor Poisoning by Self-Supervised Representation Learning
Wang, H., Karami, S., Dia, O., Ritter, H., Emamjomeh-Zadeh, E., Chen, J., Xiang, Z. J., Miller, D. J., & Kesidis, G. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10097244 · Cited by 1 · see citing works
Self-citationStrong evasive backdoor attacks and an ensemble defense
Yang, G., Kesidis, G., & Miller, D. J. (2026) · ETRI Journal · doi:10.4218/etrij.2026-0167
Self-citation
- Authors
- Xiang, Z., Miller, D. J., Wang, H., & Kesidis, G.
- Affiliations
- Pennsylvania State University (US)
- Corresponding author
- Xiang, Zhen
- Miller, David J.
- Wang, Hang
- Kesidis, George
- Venue
- Neural Computation
- Publisher
- MIT Press - Journals
- Published
- 2021-04-13
- Volume
- 33
- Issue
- 5
- Pages
- 1329-1371
- ISSN
- 0899-7667 (ISSN-L)
- 1530-888X
- OpenAlex
- W3132644649
- Cited by
- 8
- References
- 43
- Citations by year
- 2026: 1 · 2024: 3 · 2023: 1 · 2022: 2 · 2021: 1
- FWCI
- 0.62
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Neural Computation (publishedVersion)
- PubMed (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Artificial Intelligence › Security and Verification in Computing
- Keywords
- backdoor attacks, backdoor detection, data poisoning, spatial invariance, robustness
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
