Blind image—variant based authentication method.

There is a plethora of image authentication techniques in the literature. Almost all of these methods use either a form of watermarking or data hiding techniques to embed data directly in the original image or its decomposed sub-bands. Recently, new multiple watermarking methods have emerged to tackle more robust forgery challenges where a single watermark fails or insufficient. Most of these methods are directed towards medical images for obvious reasons. This paper presents a new precise blind image-variant authentication method. This method initiates a new direction for research on image authentication as it does not embed any sort of data in the original image or its decomposed sub-band(s) but rather produces an image-variant via manipulating some statistics of the original image. These variations represent two sequences which were infused in two QR-codes that enabled the variation process. We tested this method against image processing attacks including JPEG compression, resizing, and Gaussian-noise using images from several databases including the USC – SIPI, McGill Calibrated Color Image, DICOM medical image. Also, we compared this method performance with results from seven 7 published methods and found that our method scored higher performance results for robustness and imperceptibility and it regenerated the variation secret sequences with 100% accuracy in each case. Furthermore, this method capacity is large enough to prevent data-overflow. [ABSTRACT FROM AUTHOR]