Recognition of Dynamic Signatures for People Verification
Shern Yau, Dinesh Kumar · 2007
Machine based identity validation has applications such as authentication of documents, for financial transactions, and for entry into restricted space and database. The ineffectiveness of password and personal identification numbers has been demonstrated by recent explosion of frauds. This paper proposes the use of unpenned dynamic signature to validate the authentic user and related transactional instruments. A comparison of the ability of various classifiers for classifying the multi-dimensional features of the dynamic signatures is reported. The technique has been tested for single user and multi user and also when the forger is actively attempting to cheat the system. The system is able to perfectly determine the authentic user from other users when the user's signature trace is secret. The system is also able to perfectly reject forgers who may have access to the user's signature, with a 10% of the authentic user signature being classified as 'unknown'.