Pixel-Wise Tamper Detection Under Generic Blur/Sharpen Attacks
Over the last decade we faced expansion of digital imagining technologies. This expansion enabled effortless modification of digital images – leading us to face new threats and challenges on image authentication. Watermarking methods provide solution to global image authentication – at a price of introducing additional changes to the original image. Limited ability to remove these additional changes appear in lossless watermarking methods. But if any image processing has occurred, changes introduced by lossless watermarking usually become irrevocable [1, 2]. Digital signature methods keep the original image intact. An important feature for special applications, i.e. medical. However traditional binary image authentication is not always enough. In order to identify tampered and trusted regions of the image, tamper localization is required. Pixel-wise resolution is preferred in order to exactly pin-point modified pixels. Acceptance of standard image processing is also desirable. There are very few papers on digital signature tamper localization with the respect to invariance against blur/sharpen processing. Block-based tamper localization method is presented in . The paper mentions being the first hashing method that can localize image tampering. Another method with pixel-wise tamper localization is presented in . Resistance to blurring/sharpening operations is not directly mentioned in . Our analysis does not confirm the method supports such image processing operations. Partial resistance against small-intensity blur/sharpen operations may be enabled in restore capable digital signature methods, although this resistance usually achieved as “side-effect” [4, 5]. Image hashing methods  based on extracted features – edges, feature points, image moments – show quite good robustness against blurring/sharpening. However, pixel-wise tamper localization based on these methods is not possible. More works on pixel-wise tamper localization are in watermarking section. Illegal watermark structure modification or restoration of original image  allows for pixel-wise tamper localization, although global nature of blur/sharpen operations makes it hardly compatible with watermarking methods. Watermarking methods capable at least partially resist affirmative image modifications, such as blurring, are based on global image structure and loose pixel-wise tamper localization resolution. Our goal is to present digital signature method with pixel-wise tamper localization ability, capable to resist blur/sharpen procedures, as usually image blurring or sharpening does not influence essence of the image. A lot of watermarking/digital signature methods fail to implement pixel-wise resolution against limited blurring or sharpening operations considering their effect on binary matrices of the image. In this paper we present blur/sharpen resistant method for image authentication and pixel-wise tamper localization. The method is implemented on a digital signature basis as we see this basis more advantageous over watermark basis . If the same technique would be applied on watermark, it data structures may not survive malicious modifications and allowable general image processing.