Privacy Protection of Biometrics Evaluation Database – A Preliminary Study on Synthetic Biometric Database

Currently, evaluation of biometric authentication systems depends on biometric databases. Such databases usually consist of biometric data collected in cooperation with a lot of volunteers. If the number of volunteers involved in a database becomes very large and if the database is circulated from a organization to other, there arises a new threat that the privacy of the volunteers might be infringed in case of database abuse. To protect the privacy of volunteers and to build a freely circulated biometric database, we propose a new method to construct synthetic biometric examples from real examples. In this paper, first we analyze the requirement of biometric database for evaluation and show the guidelines to build a synthetic biometric database. Then we show the preliminary case study on face database. The case study includes the data distribution analysis in a PCA subspace of 330 volunteer faces. Finally, we show some examples synthesized by our method.