Image Authentication Using Robust Image Hashing with Localization and Self-Recovery

Abstract

The rapid growth of efficient tools, which generate and edit digital images demands effective methods for assuring integrity of images. A semi-fragile block-based image authentication technique is proposed which can not only localize the alteration detections but also recover the missing contents. The proposed technique distinguishes content-preserving manipulations from the content alterations using secure image hashing instead of cryptographic hashing. The original image is divided into large blocks (sub-images) which are also divided into 8×8 blocks. Secure image hashing is utilized to generate the sub-image hash (signature) which may slightly change when the content-preserving manipulations are applied. Furthermore, the sub-image code is generated using the JPEG compression scheme. Then, two sub-image hash copies and the sub-image code are embedded into relatively-distant sub-images using a doubly linked chain which prevents the vector quantization attack. The hash and code bits are robustly embedded in chosen discrete cosine transform (DCT) coefficients exploiting a property of DCT coefficients which is invariant before and after JPEG compression. The experimental results show that the proposed technique can successfully both localize and compensate the content alterations. Furthermore, it can effectively thwart many attacks such as vector quantization attacks.