Developing a Java-Based Keystroke Biometric System for Long-Text Input
Giang Ngo, Justin Simone · 2006
Java-based feature extraction and pattern classification programs were developed for a Pace University CSIS Doctorate of Professional Studies student completing her thesis on long-text-input keystroke biometrics. Although the general functionality of these programs was developed for a prior DPS researcher using the SAS programming language, the current Java-based system was developed for added functionality, increased usability, and ease of learning. To optimize the results in ideal and application-oriented conditions the programming team worked with the researcher to increase the feature set, optimize the pattern classifier, and provide options for different fallback scenarios for small-samples. The current feature extraction program has 239 features and five parameters to optimize, and the pattern classifier has two procedural modes with seven parameters to enable/disable various feature sets.