Face Detection for Privacy Protected Images

In recent years, privacy has received increasing attention to the development of the Internet. Yet, the problem still exists in the field of image processing that needs to be solved urgently. The cloud server has an abundant resource on computation for the specified computing tasks. With the rapid growth of personal and enterprise image data, more and more computing tasks are difficult to handle locally and are usually be outsourced to the cloud server for computing. Therefore, private image data protection while image processing has become our main concern. To solve this problem, we propose a scheme to extract secure Haar feature for privacy-preserving images, which can be applied to the face detection task. This paper solves the problem of extracting Haar feature from privacy-preserved images for the first time. Besides, we compare the numerical values in the ciphertext domain with double encryption protocols based on homomorphic encryption. In this scheme, we adopt Li
et al.
’s homomorphic encryption algorithm to protect image privacy. The security analysis of Li
et al.
’s encryption algorithm shows that the scheme we proposed is secure under the known plaintext attack and ciphertext attack. The experimental results show that the scheme can be implemented in polynomial time.