Secure robot face recognition in cloud environments

Abstract

In robotic applications, recognition of human faces at high speed is important to improve human-machine interaction. Recognition speed can be limited by the on-board computation capabilities of the robots. These limitations are overcome by introducing cloud technology in face recognition. The cloud supplies the required memory and processor speed dynamically, when needed. However, public clouds are shared with many users, hence chance of image database insecurity. To address this problem, we explore several possibilities to achieve robot secure face recognition where face recognition is performed on the encrypted data. The performance of eight encryption algorithms in terms of security and recognition accuracy is measured. A face recognition test is conducted in a robot and cloud environments and the average time to recognize a face through simple Principal Component Analysis (PCA) over encrypted data is measured.