Improving on-line signature skillfulness

Abstract

One of the biggest challenges in on-line signature verification is the detection of counterfeited signatures. Recently, novel schemes based on the kinematic theory of rapid human movements and its associated Sigma-Lognormal model has been proposed to improve the detection of on-line skilled forgeries. But for a more realistic and reliable estimation of the forgery detection rate, we would need more challenging on-line forgeries than those included in current databases. To get better on-line skilled forgeries, this paper aimed at leveraging the Sigma-Lognormal model to improve the skill of any online forged signature. Specifically, we propose to replace the original velocity profile of any on-line signature by a synthetic Sigma-Lognormal profile. The new profile emulates a genuine-like velocity profiles without modifying the original ballistic trajectory. Experimental results were performed with the 132 on-line users of publicly BiosecureID database. It is shown that the detection rate of forgeries is significantly worsened when the velocity profile is replaced by the synthetic one. A countermeasure to detect this kind of improved fake signatures is also proposed.