Machine Learning Security: Threats, Countermeasures, and Evaluations
Open accessThe DOI matches its Crossref record; 46 of the 46 references it cites were found in a registry or an index.
Abstract
Machine learning has been pervasively used in a wide range of applications due to its technical breakthroughs in recent years. It has demonstrated significant success in dealing with various complex problems, and shows capabilities close to humans or even beyond humans. However, recent studies show that machine learning models are vulnerable to various attacks, which will compromise the security of the models themselves and the application systems. Moreover, such attacks are stealthy due to the unexplained nature of the deep learning models. In this survey, we systematically analyze the security issues of machine learning, focusing on existing attacks on machine learning systems, corresponding defenses or secure learning techniques, and security evaluation methods. Instead of focusing on one stage or one type of attack, this paper covers all the aspects of machine learning security from the training phase to the test phase. First, the machine learning model in the presence of adversaries is presented, and the reasons why machine learning can be attacked are analyzed. Then, the machine learning security-related issues are classified into five categories: training set poisoning; backdoors in the training set; adversarial example attacks; model theft; recovery of sensitive training data. The threat models, attack approaches, and defense techniques are analyzed systematically. To demonstrate that these threats are real concerns in the physical world, we also reviewed the attacks in real-world conditions. Several suggestions on security evaluations of machine learning systems are also provided. Last, future directions for machine learning security are also presented.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Explaining and Harnessing Adversarial Examples
Goodfellow, I. J., Shlens, J., & Szegedy, C. (2014) · arXiv · doi:10.48550/arxiv.1412.6572 · Cited by 10,381
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
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
Towards Deep Learning Models Resistant to Adversarial Attacks
Madry, A., Makelov, A., Schmidt, L., Tsipras, D., & Vladu, A. (2017) · arXiv · doi:10.48550/arxiv.1706.06083 · Cited by 6,602
Adversarial examples in the physical world
Kurakin, A., Goodfellow, I. J., & Bengio, S. (2018) · Artificial Intelligence Safety and Security · doi:10.1201/9781351251389-8 · Cited by 5,375
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 · cited 1×
“After the methods which introduce perturbations to fool a network on a single image each time (e.g., FGSM [2], Deep-Fool [57]), Moosavi-Dezfooli et al. [51] propose universal adversarial perturbations which can fool a network on any image.”
MethodMembership Inference Attacks Against Machine Learning Models
Shokri, R., Stronati, M., Song, C., & Shmatikov, V. (2017) · 2017 IEEE Symposium on Security and Privacy (SP) · doi:10.1109/sp.2017.41 · Cited by 4,754 · cited 3×
“Particularly, they use the target model’s prediction of training and non-training data to train a membership inference model [70].”
BackgroundMethodThe 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
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 · cited 2×
“They use the substitute model to generate adversarial examples which are found to be effective in the target model as well [58].”
MethodUniversal 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
One Pixel Attack for Fooling Deep Neural Networks
Su, J., Vargas, D. V., & Sakurai, K. (2019) · IEEE Transactions on Evolutionary Computation · doi:10.1109/tevc.2019.2890858 · Cited by 2,406 · cited 1×
“As an extreme case, Su et al. [48] propose a method that fools the classifier by modifying just one pixel of an image.”
MethodPrivacy-preserving deep learning
Shokri, R., & Shmatikov, V. (2015) · Proceedings of the 22nd ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/2810103.2813687 · Cited by 2,356 · cited 6×
“Differential privacy is also introduced to obfuscate the parameters for protection [74].”
BackgroundMethodInfluentialAdversarial machine learning at scale
Kurakin, A., Goodfellow, I., & Bengio, S. (2016) · arXiv · doi:10.48550/arxiv.1611.01236 · Cited by 2,325
Robust physical-world attacks on deep learning visual classification
Eykholt, K., Evtimov, I., Fernandes, E., Li, B., Rahmati, A., Xiao, C., Prakash, A., Kohno, T., & Song, D. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00175 · Cited by 2,173
Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
Akhtar, N., & Mian, A. (2018) · IEEE Access · doi:10.1109/access.2018.2807385 · Cited by 2,021
SecureML: A System for Scalable Privacy-Preserving Machine Learning
Mohassel, P., & Zhang, Y. (2017) · 2017 IEEE Symposium on Security and Privacy (SP) · doi:10.1109/sp.2017.12 · Cited by 1,936 · cited 3×
“The distributed learning framework methods [74], [118] and the secure aggregation method [113] can protect the privacy of data for multiple users in distributed deep learning framework, but the encryption algorithm or security protocol used in the aggregation of multiple users’ data will increase…”
BackgroundMethodSelf-citationAdversarial Examples: Attacks and Defenses for Deep Learning
Yuan, X., He, P., Zhu, Q., & Li, X. (2019) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2018.2886017 · Cited by 1,733 · cited 3×
“Studies [1], [8] also suggest that training data incompletion is one of the reasons for the existence of adversarial examples.”
BackgroundPrivacy-Preserving Deep Learning via Additively Homomorphic Encryption
Phong, L. T., Aono, Y., Hayashi, T., Wang, L., & Moriai, S. (2018) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2017.2787987 · Cited by 1,662
How to backdoor federated learning
Bagdasaryan, E., Veit, A., Hua, Y., Estrin, D., & Shmatikov, V. (2018) · arXiv · doi:10.48550/arxiv.1807.00459 · Cited by 1,495 · cited 3×
“In these methods [12], [33]–[36], the attacker first adds the well-designed backdoor signal to the clean data to generate poisoning data, and then injects the poisoning data into the training set to re-train the target model.”
BackgroundMethodDeep Models Under the GAN: Information Leakage from Collaborative Deep Learning
Hitaj, B., Ateniese, G., & Perez-Cruz, F. (2017) · Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/3133956.3134012 · Cited by 1,377 · cited 3×
“However, the attacker in [71] needs to be an insider of the collaborative deep learning framework which has permission to obtain model parameters from the service provider.”
BackgroundInfluentialPoisoning Attacks against Support Vector Machines
Biggio, B., Nelson, B., & Laskov, P. (2012) · arXiv · doi:10.48550/arxiv.1206.6389 · Cited by 1,331
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
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
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 · cited 5×
“These methods [12], [33]–[36] are stealthy and can perform backdoor attacks without affecting the performance of the model, e.”
BackgroundMethodInfluentialPrivacy Risk in Machine Learning: Analyzing the Connection to Overfitting
Yeom, S., Giacomelli, I., Fredrikson, M., & Jha, S. (2018) · 2018 IEEE 31st Computer Security Foundations Symposium (CSF) · doi:10.1109/csf.2018.00027 · Cited by 1,001
Manipulating Machine Learning: Poisoning Attacks and Countermeasures for Regression Learning
Jagielski, M., Oprea, A., Biggio, B., Liu, C., Nita-Rotaru, C., & Li, B. (2018) · 2018 IEEE Symposium on Security and Privacy (SP) · doi:10.1109/sp.2018.00057 · Cited by 861
Stealing Machine Learning Models via Prediction APIs
Tramèr, F., Zhang, F., Juels, A., Reiter, M. K., & Ristenpart, T. (2016) · arXiv · doi:10.48550/arxiv.1609.02943 · Cited by 796
Lowd, D., & Meek, C. (2005) · Proceedings of the eleventh ACM SIGKDD international conference on Knowledge discovery in data mining · doi:10.1145/1081870.1081950 · Cited by 755 · cited 2×
“Lowd and Meek [40] introduce the adversarial learning problem, in which the adversary tries to reverse engineering the classifier through sending a number of queries.”
BackgroundTowards Poisoning of Deep Learning Algorithms with Back-gradient Optimization
Muñoz-González, L., Biggio, B., Demontis, A., Paudice, A., Wongrassamee, V., Lupu, E. C., & Roli, F. (2017) · Proceedings of the 10th ACM Workshop on Artificial Intelligence and Security · doi:10.1145/3128572.3140451 · Cited by 549 · cited 3×
“They also reverse the learning process to reduce the complexity of the attack [28].”
BackgroundInfluentialSoK: Security and Privacy in Machine Learning
Papernot, N., McDaniel, P., Sinha, A., & Wellman, M. P. (2018) · 2018 IEEE European Symposium on Security and Privacy (EuroS&P) · doi:10.1109/eurosp.2018.00035 · Cited by 502 · cited 4×
“Papernot et al. [11] systematize the security and privacy issues in machine learning.”
BackgroundInfluentialSemi-supervised Knowledge Transfer for Deep Learning from Private Training Data
Papernot, N., Abadi, M., Erlingsson, Ú., Goodfellow, I., & Talwar, K. (2016) · arXiv · doi:10.48550/arxiv.1610.05755 · Cited by 492
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
Certified Defenses for Data Poisoning Attacks
Steinhardt, J., Koh, P. W., & Liang, P. (2017) · arXiv · doi:10.48550/arxiv.1706.03691 · Cited by 479
Privacy-Preserving Machine Learning: Threats and Solutions
Al-Rubaie, M., & Chang, J. M. (2019) · IEEE Security & Privacy · doi:10.1109/msec.2018.2888775 · Cited by 444
PRADA: Protecting Against DNN Model Stealing Attacks
Juuti, M., Szyller, S., Marchal, S., & Asokan, N. (2019) · 2019 IEEE European Symposium on Security and Privacy (EuroS&P) · doi:10.1109/eurosp.2019.00044 · Cited by 440 · cited 1×
“Juuti et al. [111] propose a technique to detect DNN model stealing attacks, named PRADA.”
MethodA Survey on Security Threats and Defensive Techniques of Machine Learning: A Data Driven View
Liu, Q., Li, P., Zhao, W., Cai, W., Yu, S., & Leung, V. C. M. (2018) · IEEE Access · doi:10.1109/access.2018.2805680 · Cited by 421
ANTIDOTE: understanding and defending against poisoning of anomaly detectors
Rubinstein, B. I. P., Nelson, B., Huang, L., Joseph, A. D., Lau, S.-H., Rao, S., Taft, N., & Tygar, J. D. (2009) · Proceedings of the 9th ACM SIGCOMM conference on Internet measurement · doi:10.1145/1644893.1644895 · Cited by 331 · cited 4×
“These methods [14], [77], [78] can improve the robustness of learning algorithms and mitigate the influence of outliers on the trained model, but these methods mainly focus on binary classification problems and introduce additional overhead.”
MethodInfluentialNightmare at test time: robust learning by feature deletion
Globerson, A., & Roweis, S. (2006) · Proceedings of the 23rd international conference on Machine learning - ICML '06 · doi:10.1145/1143844.1143889 · Cited by 309
Februus: Input Purification Defense Against Trojan Attacks on Deep Neural Network Systems
Doan, B. G., Abbasnejad, E., & Ranasinghe, D. C. (2020) · Annual Computer Security Applications Conference · doi:10.1145/3427228.3427264 · Cited by 263
Liu, Y., Xie, Y., & Srivastava, A. (2017) · 2017 IEEE International Conference on Computer Design (ICCD) · doi:10.1109/iccd.2017.16 · Cited by 241
Data poisoning attacks on factorization-based collaborative filtering
Li, B., Wang, Y., Singh, A., & Vorobeychik, Y. (2016) · arXiv · doi:10.48550/arxiv.1608.08182 · Cited by 235
A uror: defending against poisoning attacks in collaborative deep learning systems
Shen, S., Tople, S., & Saxena, P. (2016) · Proceedings of the 32nd Annual Conference on Computer Security Applications · doi:10.1145/2991079.2991125 · Cited by 228 · cited 2×
“Shen et al. [83] propose a system, named AUROR, to defend the collaborative deep learning systems.”
BackgroundStronger Data Poisoning Attacks Break Data Sanitization Defenses
Koh, P. W., Steinhardt, J., & Liang, P. (2018) · arXiv · doi:10.48550/arxiv.1811.00741 · Cited by 189
Backdoor Embedding in Convolutional Neural Network Models via Invisible Perturbation
Liao, C., Zhong, H., Squicciarini, A., Zhu, S., & Miller, D. (2018) · arXiv · doi:10.48550/arxiv.1808.10307 · Cited by 110
Mitigating Poisoning Attacks on Machine Learning Models: A Data Provenance Based Approach
Baracaldo, N., Chen, B., Ludwig, H., & Safavi, J. A. (2017) · Proceedings of the 10th ACM Workshop on Artificial Intelligence and Security · doi:10.1145/3128572.3140450 · Cited by 108 · cited 1×
“These methods [23], [80]–[82] can improve the robustness of learning algorithms, but these methods are not generic and only suitable for specific algorithms.”
MethodTowards Data Poisoning Attacks in Crowd Sensing Systems
Miao, C., Li, Q., Xiao, H., Jiang, W., Huai, M., & Su, L. (2018) · Proceedings of the Eighteenth ACM International Symposium on Mobile Ad Hoc Networking and Computing · doi:10.1145/3209582.3209594 · Cited by 77
Cited by
Differential Privacy for Deep and Federated Learning: A Survey
Ouadrhiri, A. E., & Abdelhadi, A. (2022) · IEEE Access · doi:10.1109/access.2022.3151670 · Cited by 445
“All presented works are dated before July 2019, nevertheless, the period after 2019 till now, had recognized the emergence of many works especially for preserving privacy in FL. Xue et al. [46] provide a detailed explanation of the different attacks that may threaten a DL model.”
BackgroundState-of-the-Art and Research Opportunities for Next-Generation Consumer Electronics
Wu, C. K., Cheng, C.-T., Uwate, Y., Chen, G., Mumtaz, S., & Tsang, K. F. (2023) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2022.3232478 · Cited by 227
A systematic review on machine learning models for online learning and examination systems
Kaddoura, S., Popescu, D. E., & Hemanth, J. D. (2022) · PeerJ Computer Science · doi:10.7717/peerj-cs.986 · Cited by 75
Cloud Versus Edge Deployment Strategies of Real-Time Face Recognition Inference
Koubaa, A., Ammar, A., Kanhouch, A., & AlHabashi, Y. (2022) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2021.3055835 · Cited by 69
The Emerged Security and Privacy of LLM Agent: A Survey with Case Studies
He, F., Zhu, T., Ye, D., Liu, B., Zhou, W., & Yu, P. S. (2024) · arXiv · doi:10.48550/arxiv.2407.19354 · Cited by 60
Metadata
- Authors
- Xue, M., Yuan, C., Wu, H., Zhang, Y., & Liu, W.
- Venue
- IEEE Access
- Published
- 2020
- Volume
- 8
- Pages
- 74720-74742
- ISSN
- 2169-3536 (ISSN-L)
- OpenAlex
- W3016493309
- Cited by
- 194
- References
- 46
- FWCI
- 9.47
- Type
- journal article
- Language
- en
- Open access
- gold
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.