A Hybrid On-line Signature Verification System Supporting Multi-Confidential Levels Defined by Data Mining Techniques

초록

On-line signature verification is used widely for user authentication in many applications. In existing systems, one confidential level is usually used as the threshold for signature verifications. This causes high false rejection ratio (FRR) to systems in which signature verification is less important, and on the other hand, high false acceptance ratio (FAR) to systems in which signature verification is more critical. Thus applying multi-confidential levels to the signature verification is crucial to solve this problem. Also, data mining technique can be used to define multi-confidential levels appropriately. This paper proposes a hybrid on-line signature verification system supporting multi-confidential levels defined by data mining techniques.