Abstract:This paper addresses issues concerning biometric authentication based on handwritten signatures. Our research aimed to check whether a handwritten signature acquired with a mobile device can effectively verify a user's identity. We present a novel online signature verification method using coordinates of points and pressure values at each point collected with a mobile device. Convolutional neural networks are used for signature verification. In this paper, three neural network models are investigated, i.e., two self-made light SigNet and SigNetExt models and the VGG-16 model commonly used in image processing. The convolutional neural networks aim to determine whether the acquired signature sample matches the class declared by the signer. Thus, the scenario of closed set verification is performed. The effectiveness of our method was tested on signatures acquired with mobile phones. We used the subset of the multimodal database, MobiBits, that was captured using a custom-made application and consists of samples acquired from 53 people of diverse ages. The experimental results on accurate data demonstrate that developed architectures of deep neural networks can be successfully used for online handwritten signature verification. We achieved an equal error rate (EER) of 0.63% for random forgeries and 6.66% for skilled forgeries.

What problem does this paper attempt to address?

The problem this paper attempts to address is the effectiveness and accuracy of biometric authentication based on handwritten signatures on mobile devices. Specifically, the study aims to verify whether handwritten signatures captured via mobile devices can effectively confirm a user's identity. To achieve this goal, the authors propose a new online signature verification method that utilizes the coordinate points and pressure values of each point collected by mobile devices, and employs Convolutional Neural Networks (CNN) for signature verification. ### Main Contributions: 1. **Preprocessing Method**: A method is proposed to convert online signatures collected on mobile devices into a valuable form suitable for the verification system. 2. **Custom Classifier**: Two custom lightweight Convolutional Neural Network models (SigNet and SigNetExt) were developed for signature recognition on mobile devices. 3. **Performance Comparison**: A comprehensive study was conducted comparing the online signature recognition performance of commonly used pre-trained VGG-16 models with the custom SigNet and SigNetExt models. ### Research Background: - Handwritten signatures are one of the oldest and most socially accepted biometric authentication methods. - With the advancement of technology, signature capture technology has improved, forming a new branch of online handwritten signatures. - Nevertheless, differences in user writing neatness lead to significant variations in the duration, position, or shape of the same person's signature samples, complicating the recognition and verification problem. - The current level of technology makes it possible to implement artificial neural networks on cloud servers, personal computers, and even mobile phones. - However, research on verification methods that can be used on mobile phones is relatively scarce, and most technologies still require the use of specialized tablets. ### Experimental Results: - The method was tested on signature data collected on mobile phones. - A subset of the multimodal database MobiBits, containing samples from 53 people of different age groups, was used. - Experimental results show that the developed deep neural network architecture can be successfully used for online handwritten signature verification. - For random forged signatures, the Equal Error Rate (EER) reached 0.63%, and for skilled forged signatures, the EER reached 6.66%. ### Innovations: - A new lightweight Convolutional Neural Network architecture is proposed, which verifies signatures using only the coordinate points and pressure values collected by mobile devices. - This solution does not require additional measurement data, which typically needs to be collected through specialized signature recognition tablets. In summary, the main contribution of this paper is the proposal of a new method that utilizes pressure sensors and Convolutional Neural Networks on mobile devices to achieve efficient online handwritten signature verification.

Online Signature Biometrics for Mobile Devices

Handwritten Signature Verification Using Hand-Worn Devices

Offline signature verification using deep neural network with application to computer vision

Offline Handwritten Signature Recognition Using Deep Convolution Neural Network

Offline Handwritten Signature Verification Using Deep Neural Networks

A Stroke-Based RNN for Writer-Independent Online Signature Verification.

Multi-dilation Convolutional Neural Network for Automatic Handwritten Signature Verification

Mobile Based Offline Handwritten Signature Forgery Identification using Convolutional Neural Network

Offline Handwriting Signature Verification: A Transfer Learning and Feature Selection Approach

A multi-dimensional review on handwritten signature verification: strengths and gaps

Handwritten signature identification based on MobileNets model and support vector machine classifier

DeepHSV: User-Independent Offline Signature Verification Using Two-Channel CNN

mmSign: mmWave-based Few-Shot Online Handwritten Signature Verification

Vision Graph Convolutional Network for Writer-Independent Offline Signature Verification

Attention-based Multiple Siamese Networks with Primary Representation Guiding for Offline Signature Verification

Inertial-Measurement- Based Biometric Authentication of Handwritten Signature

A new algorithm for identity verification based on the analysis of a handwritten dynamic signature

Acoustic Sensing Based on Online Handwritten Signature Verification

FCN+RL: A Fully Convolutional Network followed by Refinement Layers to Offline Handwritten Signature Segmentation

Exploiting Complexity in Pen- and Touch-based Signature Biometrics

Offline Signature Verification Using A Region Based Deep Metric Learning Network