Privacy Preserving Encoder Classifier for Access Control Based on Face Recognition

Processing private data to control the access of a system by identifying users and blocking unknowns is becoming increasingly important and requires the development of a privacy-preserving approach. This article presents a privacy-preserving access control method based on facial recognition. This method highlights the performance of classifiers based on generated output. A classifier based on generated output, a ResNet encoder, was used and tested on both clear data and encrypted data. The private data, the face images, were encrypted using a new perceptual image encryption method based on the discrete cosine transform.