README.md

Comparing Anomaly-Detection Algorithms for Keystroke Dynamics

What is keystroke dynamics (or keystroke biometrics)?

• study of whether people can be distinguished by their typing rhythms.

Applications

• acting as an electronic fingerprint
• access-control and authentication mechanism
• detecting computer-based crimes

Existing work

Comparing Anomaly-Detection Algorithms for Keystroke Dynamics

• http://www.cs.cmu.edu/~maxion/pubs/KillourhyMaxion09.pdf
• http://www.cs.cmu.edu/~keystroke/

Data

Refer to DataDescription.txt

Detectors implemented and Results

Detector	Average Equal-Error Rate	Standard deviation of EER
Manhattan Scaled Detector	0.0945	0.068375
Outlier Count (z-score)	0.103167	0.07691
Nearest Neighbor (Mahalanobis)	0.1075	0.06213
SVM (one-class)	0.12068	0.0586
Manhattan Filtered	0.12535	0.081299
Mahalanobis	0.1337	0.06678
Mahalanobis Normed	0.1337	0.06678
Manhattan	0.15	0.09
K-Means	0.1559	0.072
Neural Network (auto-assoc)	0.16417	0.0914199
Euclidean	0.16929	0.0931429
Euclidean Normed	0.2107	0.1174
Neural Network (standard)	0.6551	0.1866

Files for detectors

• Neural Network (standard) - NeuratNetStandardDetector.ipynb
• Neural Network (auto-assoc) - NeuratNetAutoAssocDetector.ipynb
• Svm - svm.ipynb
• KMeans - kmeans.ipynb
• Other detectors - KeystrokeDynamics.ipynb