DeepAID: Interpreting and Improving Deep Learning-based Anomaly Detection in Security Applications
Open accessThe DOI matches its Crossref record; 78 of the 78 references it cites were found in a registry or an index.
Abstract
Unsupervised Deep Learning (DL) techniques have been widely used in various security-related anomaly detection applications, owing to the great promise of being able to detect unforeseen threats and superior performance provided by Deep Neural Networks (DNN). However, the lack of interpretability creates key barriers to the adoption of DL models in practice. Unfortunately, existing interpretation approaches are proposed for supervised learning models and/or non-security domains, which are unadaptable for unsupervised DL models and fail to satisfy special requirements in security domains.
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.
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 source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Similarly, a recent systematic review [31] highlighted that research on XAI for cybersecurity shows that many XAI applications are crafted without a thorough understanding of their integration into analyst workflows.”
From HuntGPT: Integrating Machine Learning-Based Anomaly Detection and Explainable AI with Large Language Models (LLMs) (2026) · cites this paper as [31]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Moreover, security literature frequently fails to differentiate between diverse use cases or clearly separate the roles of model users and designers, which could potentially lead to diminished adoption.”
From HuntGPT: Integrating Machine Learning-Based Anomaly Detection and Explainable AI with Large Language Models (LLMs) (2026) · cites this paper as [31]
Experimental — accuracy not yet measured
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
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 1×
“This problem can be solved by gradient-based optimization approaches (we use Adam optimizer [28] in this study) since the objective function is fully differentiable.”
MethodPyTorch: An Imperative Style, High-Performance Deep Learning Library
Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al. (2019) · arXiv · doi:10.48550/arxiv.1912.01703 · Cited by 31,473 · cited 2×
“, PyTorch [40]) set the embedding layer to be indifferentiable.”
BackgroundDistributed Representations of Words and Phrases and their Compositionality
Mikolov, T., Sutskever, I., Chen, K., Corrado, G., & Dean, J. (2013) · arXiv · doi:10.48550/arxiv.1310.4546 · Cited by 24,533 · cited 1×
“However, embeddings of inputs are derived by an indifferentiable look-up operation [34] in many cases.”
BackgroundA Unified Approach to Interpreting Model Predictions
Lundberg, S., & Lee, S.-I. (2017) · arXiv · doi:10.48550/arxiv.1705.07874 · Cited by 23,707 · cited 1×
“For example, SHAP interpretation [32] is adapted to (single) Autoencoder in [2], which cannot be applied to Kitsune (ensemble Autoencoders) or other DL models.”
Method"Why Should I Trust You?": Explaining the Predictions of Any Classifier
Ribeiro, M. T., Singh, S., & Guestrin, C. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939778 · Cited by 17,555
Gauriat, C.-M., Pencolé, Y., Ribot, P., & Brouillet, G. (2024) · HAL (Le Centre pour la Communication Scientifique Directe) · doi:10.4230/oasics.dx.2024.27 · Cited by 12,586
Chandola, V., Banerjee, A., & Kumar, V. (2009) · ACM Computing Surveys · doi:10.1145/1541880.1541882 · Cited by 11,687
Learning Deep Features for Discriminative Localization
Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., & Torralba, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.319 · Cited by 11,428 · cited 2×
“Secondly, most prior methods are designed for non-security domains such as computer vision [16, 44, 62, 65] to understand the mechanism of DNNs.”
DeepWalk: online learning of social representations
Perozzi, B., Al-Rfou, R., & Skiena, S. (2014) · Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/2623330.2623732 · Cited by 10,123 · cited 2×
“The training data is collected from the purely benign authentication graphs through DeepWalk [41], which is a NN-based GE method.”
MethodTowards 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 · cited 1×
“We eliminate the box constraint (5) through the idea of change-of-variables [8] with tanh function.”
Method
Research works printed but not linked: 2
- Alexander D. Kent. 2015. Cybersecurity Data Sources for Dynamic Network Research. In Dynamic Networks in Cybersecurity. Imperial College Press.
- Limin Yang, Wenbo Guo, Qingying Hao, Arridhana Ciptadi, Ali Ahmadzadeh, Xinyu Xing, and Gang Wang. 2021. CADE: Detecting and Explaining Concept Drift Samples for Security Applications. In 30th USENIX Security Symposium (USENIX Security).
Cited by
Items 101–123 of 123
BehaMiner: System Behavior Mining for Audit Log Based on Graph Learning
Ma, X., Liu, X., Li, C., Yu, Z., Zhang, Q., Lv, Q., Wang, Y., & Jiang, J. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-71464-1_28
Certainly Uncertain: Demystifying ML Uncertainty for Active Learning in Network Monitoring Tasks
Dietz, K., Hajizadeh, M., Wehner, N., Geißler, S., Casas, P., Seufert, M., & Hoßfeld, T. (2024) · International Conference on Network and Service Management (CNSM) · doi:10.23919/cnsm62983.2024.10814433
Fu, C., Qi, L. I., & Xu, K. (2025) · Scientia Sinica Informationis · doi:10.1360/ssi-2025-0025
End-to-End Attack Scene Reconstruction in a Host With Rules and Anomaly-Based Detection Models
Wang, S., Sun, H., Wang, Z., Zhou, T., Yin, X., Han, D., Zhang, H., Shi, X., & Yang, J. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3588251 · Cited by 1
Self-citationAI-Driven Multi-Agent System for Real-Time Security Analysis of Software Releases
Fedorov, S. (2025) · International Conference on Intelligent Computing and Robotics (ICICR) · doi:10.1109/icicr65456.2025.00212
“Studies indicate that security architectures leveraging MAS principles surpass traditional monolithic models by decentralizing security tasks among dedicated agents, which helps alleviate computational overhead and improves threat detection accuracy [7].”
An Approach to Anomaly Detection with Dynamic Threshold Definition for Real-World Environments
Marques, G. C., de Souza, M. D., Santos, F., Zanchettin, C., & Nováis, P. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-08465-1_13
Behavior Molecule-Level Learning by Behavioral Interior Reconstruction
Wang, C. (2025) · Springer eBooks · doi:10.1007/978-981-95-1013-9_3
Towards Automatic Rule Extraction for Intrusion Detection with Explainable AI
Wang, X., Miao, H., Li, Z., Wang, W., & Liu, F. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11228580 · Cited by 1
“DeepAID [9] introduced a framework for unsupervised learning, enhancing the transparency of deep learning anomaly detection models, though its complexity and adaptability to specific architectures remain challenges.”
Guntur, T. R., Grimes, K. P., Velnambi, N., Alayed, W. M., & Abdallah, M. (2025) · IEEE Access · doi:10.1109/access.2025.3639362
“We refer to prior works such as [36] and [48], as well as established XAI systems like DeepAID [49] and Kitsune [50], which demonstrate how explainability can be layered onto IDS pipelines.”
Interpretable Defense Against Structural Adversarial Attacks on Android Malware Detection
Wei, W., Zhao, K., Zhou, H., Li, J., Wu, S., Fan, M., Luo, X., Wang, T., Zhou, K., Liu, T., et al. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3639962 · Cited by 1
“Han et al. [69] proposed a general framework DeepAID for interpreting and improving unsupervised deep learning models that aim at anomaly detection in security applications.”
Detecting and Characterizing APT Attacks in the Open World
Xi, H., Han, Y., Li, X., Song, J., Zhang, Z., Hai, W. A. N., & Zhao, X. (2025) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads67057.2025.11323077
“Recent decades have witnessed the significant success of deep learning methods supported by arti-ficial neural networks in various fields owing to their excellent feature representation and modeling capabilities, especially in classification [10], [13], [30], [35].”
Deep Learning for Adaptive Intrusion Detection in Open-World: A Survey
Wang, R., Zhuang, H., Cao, H., & Zheng, Y. (2025) · IEEE International Conference on Data Science in Cyberspace · doi:10.1109/dsc67331.2025.00018
“Deep neural networks are inherently sensitive to feature perturbations; even slight input changes can cause incorrect predictions, posing risks in safety-critical scenarios [27].”
MEB: A Backdoor Detection Framework for Pre-training Based Malicious Traffic Detection
Wang, K., Zhou, J., Yang, L., Li, W., Jie, Z., Jiang, Z., & Wang, Z. (2025) · IEEE International Conference on Trust, Security and Privacy in Computing and Communications (TrustCom) · doi:10.1109/trustcom66490.2025.00043
“They extract important features by approximating the decision boundary [40], and searching reference points [13].”
InfluentialSelf-citationYou, P., Chen, P., Xia, Y., Li, X., Bian, A., Zeng, S., & Gao, H. (2026) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-032-21171-2_30
Qi, R., Le, W., Wang, R., Xiang, Y., Yao, Y., & Jiang, Z. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-29921-5_18
Beyond Token Attention: Contiguous and Interpretable Payload Identification for Web Attacks
Yang, Y., Zhao, J. Z., Pei, C., Wan, W., Long, C., Du, G., Liu, Y., & Li, J. (2026) · International Conference on Dependable Systems & Networks/Proceedings · doi:10.1109/dsn69566.2026.00044
“Similarly, Han et al. [32] leverage expert feedback via an interpreter–distiller design, while He et al. [6] introduce Finer, a self-supervised interpreter improving interpretation fidelity through attribution learning.”
Detection and Interpretation of Malicious Network Traffic via a Novel 8-Channel Image Representation
Wang, Z., Luo, J., Teng, S., & Yuan, S. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3012-9_12
Self-citationZhang, H., Liu, X., Qian, L., Liu, G. M., Zhang, T., Zhao, K., Tong, Y., Xiao, Z., He, D., Li, Y., et al. (2026) · ACM SIGCOMM Conference (SIGCOMM) · doi:10.1145/3789240.3829107
Self-citationExplainable AI and few-shot prompting in an LLM-based agent for network intrusion detection systems
Labed, A., Saadaoui, I., Kazar, O., & Li, Z. (2026) · Cluster Computing · doi:10.1007/s10586-026-06484-0
CASTrace: Integrating causal knowledge analysis and sequential learning for APT attack investigation
Cheng, W., Weng, Z., Wang, Z., Zhang, Y., Zhu, T., Lv, M., Lin, Y., Chen, H., & Xia, Z. (2026) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2026.103990
Self-citationManai, E., Salhab, W., Mejri, M., & Fattahi, J. (2026) · IEEE Access · doi:10.1109/access.2026.3733220
Hsu, T.-C., & Lin, S.-C. (2026) · Applied Sciences · doi:10.3390/app162010172
Metadata
- Authors
- Han, D., Wang, Z., Chen, W., Zhong, Y., Wang, S., Zhang, H., Yang, J., Shi, X., & Yin, X.
- Affiliations
- Tsinghua University (CN)
- ORCID
- Venue
- Proceedings of the 2021 ACM SIGSAC Conference on Computer and Communications Security
- Publisher
- ACM
- Published
- 2021-11-12
- Pages
- 3197-3217
- ISSN
- 1543-7221 (ISSN-L)
- OpenAlex
- W3200583622
- Cited by
- 121
- References
- 76
- Citations by year
- 2026: 9 · 2025: 38 · 2024: 42 · 2023: 25 · 2022: 7
- FWCI
- 8.53
- Type
- conference paper
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- ACM SIGSAC Conference on Computer and Communications Security (CCS) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Key Research and Development Program of China (2018YFB1800200)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Anomaly Detection Techniques and Applications
- Keywords
- anomaly detection, interpretability, unsupervised deep learning, deep neural network, security applications, explainable AI
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.