Comparing anomaly-detection algorithms for keystroke dynamics
The DOI matches its Crossref record; 20 of the 20 references it cites were found in a registry or an index.
Abstract
Keystroke dynamics-the analysis of typing rhythms to discriminate among users-has been proposed for detecting impostors (i.e., both insiders and external attackers). Since many anomaly-detection algorithms have been proposed for this task, it is natural to ask which are the top performers (e.g., to identify promising research directions). Unfortunately, we cannot conduct a sound comparison of detectors using the results in the literature because evaluation conditions are inconsistent across studies. Our objective is to collect a keystroke-dynamics data set, to develop a repeatable evaluation procedure, and to measure the performance of a range of detectors so that the results can be compared soundly. We collected data from 51 subjects typing 400 passwords each, and we implemented and evaluated 14 detectors from the keystroke-dynamics and pattern-recognition literature. The three top-performing detectors achieve equal-error rates between 9.6% and 10.2%. The results-along with the shared data and evaluation methodology-constitute a benchmark for comparing detectors and measuring progress.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
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?
In the citing paper
“The keystroke dynamics benchmark introduced by CMU [24] established a standard evaluation protocol based on hold time and inter-key latency features, showing that temporal statistics can uniquely identify users and detect impostors.”
From A SIEM-Integrated Cybersecurity Prototype for Insider Threat Anomaly Detection Using Enterprise Logs and Behavioural Biometrics (2025) · cites this paper as [30]
Citations
Cites
(2012) · Encyclopedia of the Sciences of Learning · doi:10.1007/978-1-4419-1428-6_5195 · Cited by 19,370
Simultaneous Statistical Inference
Miller, R. G. (1981) · Springer Series in Statistics · doi:10.1007/978-1-4613-8122-8 · Cited by 5,604
Handbook of Parametric and Nonparametric Statistical Procedures
(2004) · Technometrics · doi:10.1198/tech.2004.s209 · Cited by 4,661
Abe, S. (2001) · Springer eBooks · doi:10.1007/978-1-4471-0285-4 · Cited by 3,721 · cited 4×
“This classic anomaly-detection algorithm [5] resembles the Euclidean and Manhattan detectors but the distance measure is more complex.”
BackgroundMethodInfluential(1982) · Cited by 1,288
Identity authentication based on keystroke latencies
Joyce, R., & Gupta, G. (1990) · Communications of the ACM · doi:10.1145/75577.75582 · Cited by 534 · cited 4×
“This detector was described by Joyce and Gupta [ 10 ] .”
MethodInfluentialThe Handbook of Parametric and Nonparametric Statistical Procedures
Kemp, F. (2003) · Journal of the Royal Statistical Society: Series D (The Statistician) · doi:10.1111/1467-9884.t01-9-00356 · Cited by 481
Authentication by Keystroke Timing: Some Preliminary Results
Gaines, R. S., Lisowski, W., Press, S. J., & Shapiro, N. Z. (1980) · Defense Technical Information Center (DTIC) · Cited by 347
Computer-access security systems using keystroke dynamics
Bleha, S., Slivinsky, C., & Hussien, B. (1990) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/34.62613 · Cited by 312 · cited 4×
“1 Joyce & Gupta (1990) [10] Manhattan (filtered) N/A 8 2 Bleha et al. (1990) [ 2 ] Euclidean (normed) 11‐17 30”
BackgroundMethodInfluentialTyping patterns: a key to user identification
Peacock, A., Ke, X., & Wilkerson, M. (2004) · IEEE Security & Privacy Magazine · doi:10.1109/msp.2004.89 · Cited by 258 · cited 1×
“Peacock et al. [17] conducted an extensive survey of the keystroke-dynamics literature .”
BackgroundUser authentication through typing biometrics features
Araujo, L. C. F., Sucupira, L. H. R., Lizarraga, M. G., Ling, L. L., & Yabu-Uti, J. B. T. (2005) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2004.839903 · Cited by 192
User authentication through typing biometrics features
Araujo, L. C. F., Sucupira, L. H. R., Lizarraga, M. G., Ling, L. L., & Yabu-uti, J. B. T. (2005) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2004.839903(410 · Cited by 189
Web-Based Keystroke Dynamics Identity Verification Using Neural Network
Cho, S., Han, C., Han, D. H., & Kim, H.-I. (2000) · Journal of Organizational Computing and Electronic Commerce · doi:10.1207/s15327744joce1004_07 · Cited by 181 · cited 2×
“To illustrate the problem we encounter when we try to use the literature to determine which detector has the best performance, suppose we tried to compare two detectors: the Neural Network (auto-assoc) detector developed by Cho et al. [4], and the Outlier-Count (z-score) detector designed by Haider et al. [8].”
MethodA multi-technique approach for user identification through keystroke dynamics
Haider, S., Abbas, A., & Zaidi, A. K. (2002) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/icsmc.2000.886039 · Cited by 119
GA-SVM wrapper approach for feature subset selection in keystroke dynamics identity verification
Yu, E., & Cho, S. (2004) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2003.1223761 · Cited by 104 · cited 7×
“Choosing passwords for a keystroke-dynamics evaluation is tricky.”
BackgroundMethodResultInfluentialContinual Retraining of Keystroke Dynamics Based Authenticator
Kang, P., Hwang, S.-S., & Cho, S. (2007) · Lecture notes in computer science · doi:10.1007/978-3-540-74549-5_125 · Cited by 93 · cited 3×
“To calculate the equal-error rate, the threshold is chosen so that the detector’s miss and false-alarm rates are equal. This measure was used by Kang et al. [ 11 ].”
MethodInfluentialPersonal Attributes Authentication Techniques.
Forsen, G. E., Nelson, M., & Staron, R. J. (1977) · Defense Technical Information Center (DTIC) · Cited by 68
Retraining a keystroke dynamics-based authenticator with impostor patterns
Lee, H., & Cho, S. (2007) · Computers & Security · doi:10.1016/j.cose.2006.11.006 · Cited by 52 · cited 1×
“Since there is no commonly accepted method for tuning the parameters of anomaly detectors for a data set without introducing bias in the evaluation results [ 12 ], we used the parameters specified by the source study when possible, and explained what parameters we used.”
MethodEqualization of keystroke timing histograms for improved identification performance
Montalvão, J., Almeida, C. A. S., & Freire, E. O. (2006) · International Telecommunications Symposium (ITS) · doi:10.1109/its.2006.4433337 · Cited by 26 · cited 1×
“[15].”
Characteristics of auto-associative MLP as a novelty detector
Hwang, B.-H., & Cho, S. (2003) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.1999.836051 · Cited by 21 · cited 1×
“” It also incorporates a feed-forward neural-network trained with the back-propagation algorithm, but unlike a typical neural network, the structure of the network is designed for use in an anomaly detector [9].”
Method
Research works printed but not linked: 3
- 2008, R A Language and Environment for Statistical Computing
- 2008, PC Tools Security guide for windows - random password generator
- 2008, Microsoft Password checker
Cited by
A survey of intrusion detection techniques for cyber-physical systems
Mitchell, R., & Chen, I.-R. (2014) · ACM Computing Surveys · doi:10.1145/2542049 · Cited by 768
HMOG: New Behavioral Biometric Features for Continuous Authentication of Smartphone Users
Sitova, Z., Sedenka, J., Yang, Q., Peng, G., Zhou, G., Gasti, P., & Balagani, K. S. (2016) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2015.2506542 · Cited by 398
“For instance, SM and SVM were top performers in a study on keystroke authentication of desktop users by Killourhy and Maxion [20].”
Biometric Authentication and Identification Using Keystroke Dynamics: A Survey
Banerjee, S. P., & Woodard, D. (2012) · Journal of Pattern Recognition Research · doi:10.13176/11.427 · Cited by 369
“Killourhy & Maxion [91] 1,3 Manhattan() S C 51 20400 - 9.”
BackgroundInfluentialA Survey of Keystroke Dynamics Biometrics
Teh, P. S., Teoh, A. B. J., & Yue, S. (2013) · The Scientific World Journal · doi:10.1155/2013/408280 · Cited by 329
“Common measure used to compute distance score introduced in the literature included but is not limited to Euclidean [77, 120, 121], Manhattan [99, 122, 123], Bhattacharyya [81, 124], Mahalanobis [125], degree of disorder [43, 76, 126], and direction similarity measure [3]. to identify and classify pattern and make cor…”
MethodYou Are How You Touch: User Verification on Smartphones via Tapping Behaviors
Zheng, N., Bai, K., Huang, H., & Wang, H. (2014) · 2014 IEEE 22nd International Conference on Network Protocols · doi:10.1109/icnp.2014.43 · Cited by 290
“Keystroke dynamics, as one of the behavioral biometrics, has been extensively studied in distinguishing users by the way they type on a computer keyboard [3, 15, 24, 28].”
Background
Metadata
- Authors
- Killourhy, K. S., & Maxion, R. A.
- Affiliations
- Carnegie Mellon University (US)
- ORCID
- Maxion, Roy A.
- Venue
- 2009 IEEE/IFIP International Conference on Dependable Systems & Networks
- Publisher
- IEEE
- Published
- 2009-06
- Pages
- 125-134
- ISSN
- 1530-0889 (ISSN-L)
- OpenAlex
- W2107589078
- Cited by
- 566
- References
- 19
- Citations by year
- 2026: 37 · 2025: 37 · 2024: 32 · 2023: 30 · 2022: 34 · 2021: 34 · 2020: 27 · 2019: 48 · 2018: 39 · 2017: 37 · 2016: 41 · 2015: 47 · 2014: 44 · 2013: 27 · 2012: 35
- FWCI
- 12.93
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- International Conference on Dependable Systems & Networks/Proceedings (publishedVersion)
- Funders
- National Science Foundation
- Carnegie Mellon University
- Army Research Office
- Topics
- Physical Sciences › Computer Science › Information Systems › User Authentication and Security Systems
- Physical Sciences › Computer Science › Artificial Intelligence › Anomaly Detection Techniques and Applications
- Physical Sciences › Computer Science › Signal Processing › Biometric Identification and Security
- Keywords
- keystroke dynamics, anomaly detection, impostor detection, user authentication, biometrics, benchmark dataset, evaluation methodology, insider attacks, equal error rate
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.