SynSig2Vec: Learning Representations from Synthetic Dynamic Signatures for Real-world Verification
This addresses the challenge of acquiring skilled forgeries for training in signature verification systems, offering a practical solution for security applications.
The paper tackles the problem of skilled forgery attacks in automatic signature verification by learning dynamic signature representations from synthetic data, achieving state-of-the-art performance on two public benchmarks without requiring real forgeries for training.
An open research problem in automatic signature verification is the skilled forgery attacks. However, the skilled forgeries are very difficult to acquire for representation learning. To tackle this issue, this paper proposes to learn dynamic signature representations through ranking synthesized signatures. First, a neuromotor inspired signature synthesis method is proposed to synthesize signatures with different distortion levels for any template signature. Then, given the templates, we construct a lightweight one-dimensional convolutional network to learn to rank the synthesized samples, and directly optimize the average precision of the ranking to exploit relative and fine-grained signature similarities. Finally, after training, fixed-length representations can be extracted from dynamic signatures of variable lengths for verification. One highlight of our method is that it requires neither skilled nor random forgeries for training, yet it surpasses the state-of-the-art by a large margin on two public benchmarks.