Practical Construction of Face-Based Authentication Systems with Template Protection Using Secure Sketch

Modern mobile devices (e.g. laptops, mobile phones, etc.) equipped with input sensors open a convenient way to do authentication by using biometrics. However, if these devices are lost or stolen, the owners will confront a highly impacted threat: their stored biometric templates, either in raw or transformed forms, can be extracted and used illegally by others. In this paper, we propose some concrete constructions of face-based authentication systems in which the stored templates are protected by applying a cryptographic technique called secure sketch. We also suggest a simple fusion method for combining these authentication techniques to improve the overall accuracy. Finally, we evaluate accuracy rates among these constructions and the fusion method with some existing datasets.