Invertible chaotic fragile watermarking for robust image authentication

Abstract

Fragile watermarking is a popular method for image authentication. In such schemes, a fragile signal that is sensitive to manipulations is embedded in the image, so that it becomes undetectable after any modification of the original work. Most algorithms focus either on the ability to retrieve the original work after watermark detection (invertibility) or on detecting which image parts have been altered (localization). Furthermore, the majority of fragile watermarking schemes suffer from robustness flaws. We propose a new technique that combines localization and invertibility. Moreover, watermark dependency on the original image and the non-linear watermark embedding procedure guarantees that no malicious attacks will manage to create information leaks.