Analysis of feature selection on the performance of multimodal keystroke dynamics biometric systems

• Autor(a) principal: Mendes, Brenda Vasiljevic Souza
• Data de Publicação: 2017
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: eng
• Título da fonte: Repositório Institucional da UFRN
• Texto Completo: https://repositorio.ufrn.br/handle/123456789/34253
• topic: keyboard keystroke dynamics; touch keystroke dynamics; biometrics; feature selection; classification accuracy

New security systems, methods or techniques need to have their performance evaluated in conditions that closely resemble a real-life situation. Moreover, biometric systems need a realistic set of biometrics data to test their accuracy when classifying individuals between legitimate users or impostors. The use of similar modalities may influence the use of the same features, however, there is no indication that basic biometrics will perform well using the same set of features. This report aims to be the first to investigate the impact of feature selection in two similar yet different biometric modalities: keyboard keystroke dynamics and touchscreen keystroke dynamics. We have found that an efficient feature selection method, chosen to suit the needs of the classification algorithm employed by the system, can multiply accuracy rates while diminishing the number of features to be processed to a small subset - which also improves the system’s processing time and overall usability.