Vector Quantization Based Semifragile Watermarking Scheme For Image Authentication

Abstract

The world is getting digitised at a pace as never before. The importance of newlinedigital images is increasing with every passing day. The images transacted for various newlinereasons are to be thus protected from forced or unforced changes in the content. With newlinethe increased exchanges of data, data security is the topmost concern of each newlineorganisation or individual. With lot of advances in image editing software, it has newlinebecome very easy to manipulate digital images. Therefore, image authentication newlinebecomes a significant research topic. Most of the methods suggested in by the newlineerstwhile researchers need improvement in issues related to watermarked image newlinequality, identification of attacks as malicious or content preserving manipulation, and newlineefficiency of the system in authentication of images. Thus, in the present work, a newlinenovel image watermarking algorithm using vector quantization (VQ) approach is newlinepresented for digital image authentication. Watermarks are embedded in two newlinesuccessive stages for image integrity verification and authentication. In the first stage, newlinea key based approach is used to embed robust zero level watermark using properties of newlineindices of vector quantized image. In the second stage, semifragile watermark is newlineembedded by using modified index key based (MIKB) method. Random keys are used newlineto improve the integrity and security of the designed system. Further, to classify an newlineattack quantitatively as acceptable or as a malicious attack, pixel neighbourhood newlineclustering approach is introduced. newlineProposed approach is evaluated on 500 images using performance measures newlinesuch as peak signal to noise ratio (PSNR) and normalized hamming similarity (NHS). newlineThe experimental results shows that proposed approach achieve average false positive newlinerate 0.00024 and the average false negative rate 0.0012. is 42 dB and 44 dB for block newlinesize 4×4 and 2×2 respectively. Reported tamper detection/localization accuracy is newline99.8%. The proposed model has been found to be robust to common content newlinepreserving attacks while fragile to content altering attack