Writer identification using online handwritten passwords

Resumen

The aim of the thesis is to investigate the enhanced safety of authentication systems using online handwritten password. In state of the art we find some online and many offline text and signature verification and recognition systems, but no system that use a handwritten password for identification. Therefore, in this thesis, it will be investigated to increase safety with an online handwritten password for authentication systems. The improvement of results in writer recognition by increasing the safety of handwritten password and exploring the possibilities of writer recognition of short texts, such as passwords under practical conditions shall be examined. For this purpose, both available public databases as well as own databases with handwritten datasets will be used / produced as research basis. As input device a device with a touch screen display (smart phone, tablet) shall be used. The preprocessing of the data, extracting the features and the classification is investigated and applied. Research for a suitable segmentation as a function of the data format will be examined. The usefulness of new features will be tested in the databases considering different standard machine learning –based methods for feature selection and/or classification. Not only theoretical investigation with popular data mining tools (Matlab, WEKA) will be applied. The proposed system will be tested in a real-world application using a prototype that will be developed for handwriting password identification and writer identification by an own dataset.