A Digital Identity Authentication Framework via BBS+ Signatures and Convolutional Neural Networks

With the advent of the digital era, the security and reliability of identity authentication technologies face unprecedented challenges. This paper proposes a digital identity authentication framework that integrates BBS+ signatures with convolutional neural networks, combining the security of cryptographic signatures with the pattern recognition capabilities of deep learning. Through constructing a signature verification model based on Triplet Network, precise recognition and verification of handwritten signatures is achieved. The experiment uses the UTSig dataset for validation, containing four types of samples: genuine signatures, skilled forgeries, simple forgeries, and opposite-hand signatures. The proposed model achieves convergence within 18 training epochs, with Triplet Loss decreasing from an initial 0.080 to 0.008, and validation accuracy reaching 94.726%. The research results demonstrate that this framework can effectively distinguish between genuine and forged signatures, providing a secure and efficient solution for digital identity authentication.