Support Vector Machines (SVM)-based biometric watermarking for offline handwritten signature

Biometric watermarking refers to the incorporation of biometrics and watermarking technology. In this paper, we present a novel biometric watermarking scheme to embed handwritten signature in the host as a notice of legitimate ownership. The core of the proposed method is the synergistic integration of a statistical classifier, i.e. the Support Vector Machine, with biometric watermarking to precisely extract the signature code from the host. We abbreviate the proposed method as SVM-BW. The performance of SVM-BW is validated against simulated frequency and geometric attacks, which include JPG compression, low pass filtering, median filtering, noise addition, scaling, rotation and cropping. Experiment results reveal that SVM-BW is able to endure severe degradation on the host fidelity. Furthermore, SVM-BW shows remarkable robustness even if the host is deliberately distorted.