INFORMATION HIDING IN ENCRYPTED IMAGES FOR PRIVACY-PRESERVATION AND COVERT COMMUNICATION

Abstract

The adoption of digital data have accentuated the importance of secure communication, pushing the advancements in data hiding techniques. Reversible Data Hiding in Encrypted Images (RDHEI) stands out as a important domain, addressing the dual requirements of robust security of cover image and data integrity. This thesis embarks on an exploitative journey through the RDHEI domain, identifying key challenges and proposing innovative solutions that mark significant strides in this domain. The thesis first proposes a groundbreaking method that introduces a novel link chain driven RDHEI method. The proposed method leverages neighborhood similarity to create variable sized link chains for high-capacity data embedding. This innovative approach introduces a novel pixel collection strategy, enhances embedding space creation, ensures reversibility within the encrypted domain, and provides flexibility in balancing error correction and image distortion. In our second work, a high-capacity RDHEI method based on difference image transfigura tion is introduced. The proposed method utilizes advanced image processing techniques such as the Median Edge Detector and difference predictor to generate images with lower magni tude prediction values, creating space for data embedding. Additionally, this method introduces an innovative cleaning process, novel difference location techniques, and Huffman encoding for weight matrix size reduction, contributing to enhanced security, direct data embedding, im proved efficiency, and the advantage of bit-plane-wise image pre-processing. The third work introduces a novel multi-level blocking and quad-tree encoding based RD HEI method, which addresses the limitations of variable block size approaches. This technique employs a quad-tree-based bit-plane encoding technique and an inter bit-plane redundancy mit igation strategy, effectively compressing the final bit stream by removing redundancies across bit-planes. The multi-level blocking RDHEI method demonstrates superior embedding capac ity while maintaining flawless original image reconstruction, introducing a novel high-capacity RDHEI technique and effectively utilizing inter and intra bit-plane redundancies. i The fourth work proposes a novel RDHEI method that employs a block-level division ap proach, innovating with 2×2 non-overlapping blocks to enhance spatial correlation exploitation for improved embedding rates and guaranteed reversibility. This method introduces a novel block-shifting mechanism, enabling quicker and more secure RDHEI without the need for a block-label map, thereby enhancing both security and operational efficiency. It also introduces an index-based approach for seamless integration of hidden data within image content, avoiding additional overhead or keys. The fifth work introduces an Adaptive Two-Stage RDHEI method which utilizes prediction error expansion and pixel value ordering to overcome the limitations of overflow, underflow, and limited embedding capacity. This method combines image encryption and RDHEI methods, enhancing security and adaptability while maintaining visual security and correlation through block rotation. It demonstrates increased embedding capacity compared to existing methods and provides flexible and confidential recovery of data and images based on key availability.