A Privacy-Preserving Handwritten Signature Verification Method Using Combinational Features and Secure KNN

The handwritten signature is a widely accepted biometric trait for the individual authentication. This paper proposes a secure and dynamic signature verification method which applies to the mobile phone. A key point for signature verification is to extract features of good distinguishability. There are four main steps in our feature extraction process, including preprocessing, attribute generation, attribute truncation and quantization, and feature generation. In addition to the global features that are extracted from the whole signature, we divide a signature into several segments and extract features from these separate regions. The final feature vector is the combination of global and regional features. A user template is constructed by averaging the feature vectors whose elements are scaled by the feature-specific factors. Then, the similarity score of the test signature to the user template can be measured by Euclidean distance. In order to protect the user privacy in a cloud computing scenario, the template and features are further protected by secure kNN which has no influence on signature matching. The performance of the proposed method is demonstrated on the SG-NOTE database acquired by Samsung Galaxy Note and the MCYT-100 database captured by a WACOM pen tablet.