SECURE SEMI-FRAGILE WATERMARKING: ENHANCING TAMPER DETECTION, LOCALIZATION, AND RECOVERY WITH CRYPTOGRAPHIC ALGORITHMS

Abstract

The use of digitized medical imagery, rich with patient health information, significantly aids in diagnostic procedures. However, the integrity of these images is paramount, as even slight variations can lead to misdiagnosis and adversely affect subsequent treatment by healthcare practitioners. Consequently, robust protection mechanisms are essential to guard against intentional manipulations such as compression, filtering, or forgery, as well as unintentional tampering and other forms of attack, thereby preserving the authenticity of these critical images. This study details a semi-fragile image watermarking methodology that employs a hash function-derived watermark for application to 512×512 grayscale host medical images. The method begins by dividing the host image into two separate regions: the Region of Interest (ROI) and the Region of Non-Interest (RONI). A fragile watermark is produced by performing the Lifting Wavelet Transform (LWT) on the RONI. This watermark is then embedded within the RONI of an image that has already been robustly watermarked, using the Least Significant Bit (LSB) replacement method.. Upon receipt, the authenticity of the watermarked image is validated, and any attacks are identified through a tamper detection process. Furthermore, a cryptosystem secret key is derived from specific coefficients using the Hybrid Chaotic Magic Transform (HCMT). The watermark itself is computed within blocks by permuting the six Most Significant Bits (MSBs) of each pixel. For every block, the Discrete Wavelet Transform (DWT) is applied to extract DC coefficients. Tampering or forgery is detected by comparing these extracted keys with the generated keys. For watermark localization, the arithmetic mean of a designated block and the Maximum Pixel Intensity (MPI) within that block are utilized. Additionally, the Lempel-Ziv-Welch (LZW) algorithm is implemented to compress the recovery data pertaining to the host image's ROI. The efficacy of this proposed method is systematically evaluated using Matlab software. The findings demonstrate that the proposed work achieves high accuracy in tampering detection, maintains good imperceptibility of the watermark, and exhibits robustness against a variety of watermarking attacks. The scheme's robustness and fragility characteristics are quantified using Normalized Cross- Correlation (NC), Bit Error Rate (BER), and Peak Signal-to-Noise Ratio (PSNR). As a result, the developed system can accurately localize and detect tampering incidents and is also capable of recovering the tampered regions of the images.