Efficient Region of Interest Encryption Based on a New Chaotic Map

Abstract

The recent advances in digital imaging and transmission over network demand a secure and efficient image encryption system. Traditional cryptographic algorithms encrypt full image, however only some part of the image is important and needs complex encryption algorithm to provide security. To address this issue, we have proposed a novel encryption technique based on YOLOv2 object detection and proposed Gauss-Square Cosine (GSC) chaotic map. YOLOv2 is used to divide the image into region of interest (ROI) and noncritical region. SHA2-256 hash of scrambled noncritical region is then used as an initial value of GSC chaotic map to encrypt ROI. ROI encryption technique uses pixel permutation to decorrelate adjacent pixels and random substitution to change pixel values. Differential and statistical analysis demonstrate the complexity of proposed chaotic map and robustness of proposed technique against various security attacks. Thus the proposed encryption technique is secure and efficient for large-sized images.