From Signatures to Graphs

Abstract

Handwritten signatures are an important authentication measure in our everyday but especially in our business activities since the fourth century (Fillingham, 1997). A signature is a sign of trust between two parties, a sign of trust that also can be abused. Forgery of signatures is therefore as old as signatures themselves. Consequently, signature verification aims at the classification of forged from genuine signatures in order to mitigate this risk. This task remained manually until the first publication of a computer-aided expert system in 1977 (Herbst and Liu, 1977; Nagel and Rosenfeld, 1977). Signature verification is a process consisting of four activities, viz. the data acquisition, the preprocessing, the feature extraction and the classification (Impedovo and Pirlo,2008). During the data acquisition signatures are either acquired offline or online. In caseof offline acquisition, signatures are available on paper and are posterior digitalised via scanner. In case of online acquisition, signatures are directly acquired via electronic input device (e.g. digital pen). The preprocessing generally aims at enhancing the previously acquired signatures in order to better extract the meaningful characteristic of a signature during the feature extraction. Finally, based on the extracted features, a signature is classified as genuine or forged by means of a particular classification method....