A New Smart Glasses-Based Dual Authentication System via Signature Image Recognition Handwriting Behaviors Identification
Xiangyu Su, Siyu Xiong, Bo Wu · 2024
As technology advances, the methods of forging handwritten signatures have become difficult to detect, posing a serious threat to personal privacy and information security. Traditional single factor authentication methods are increasingly unable to withstand these security challenges. On the other hand, in recent years, with the development of computer vision-based authentication methods, contactless authentication has become a necessary option, particularly in security and authentication. This study leverages the system and cameras of smart glasses as a means of computer vision to develop a dual authentication system. The system utilizes a camera of smart glasses to gather signature data and handwriting behavior data. It employs Siamese neural networks for signature feature extraction and MediaPipe for handwriting behavior feature extraction. The data is then deep learned using 2D CNN and LSTM neural networks for effective training and authentication, respectively. Experimental results show that the system achieves an accuracy of 82 % in signature authentication and an accuracy of 75% in handwriting behavior authentication. This demonstrates that by combining the dual authentication methods of signature and handwriting behavior, the system can significantly enhance the security of personal identity verification and mitigate the security risks associated with forged signatures.