Statical algorithms for identifying unique features from a person's handwritten signature
Abstract
Statical algorithms for identifying unique features from a person's handwritten signature
Incoming article date: 13.02.2024One of the most reliable methods of identity verification are biometric authentication methods. There are two types of methods: static and dynamic. Static methods include fingerprint scanning, 3D facial recognition, vein patterns, retina scanning, etc. Dynamic methods include voice verification, keyboard handwriting and signature recognition. As of today, static methods have the lowest type I and II error rates, because their primary principle of operation is based on capturing a person's biometric characteristics, which do not change throughout their lifetime. Unfortunately, this advantage, which accounts for such low type I and II error rates, is also a drawback when implementing this method for widespread use among internet services. If biometric data is compromised, user can no longer safely use method everywhere. Dynamic biometric authentication methods are based on a person's behavioral characteristics, allowing user to control information entered for authentication. However, behavioral characteristics are more vulnerable to changes than static, resulting in significantly different type I and II errors. The aim of this work is to analyze one of the dynamic methods of biometric authentication, which can be used in most internal and external information systems as a tool for authorization or confirmation of user intentions. Biometric user authentication based on their handwritten signature relies on comparing unique biometric features that can be extracted from signature image. These unique features are divided into two categories: static and dynamic. Static features are extracted from signature image, based on characteristics such as point coordinates, total length, and width of the signature. Dynamic features are based on coordinate dependency of the signature points over time. More unique features are identified and more accurately each is weighted, the better type I and II error rates will be. This work focuses on algorithms that extract unique features from static characteristics of signature, as most signature peculiarities are identified from the dependencies of writing individual segments of the signature image.
Keywords: static algorithms, metrics, signature length, scaling, signature angle