Compressed-image authentication using compensated signature embedding

In this paper, an effective content-based authentication system with localization is proposed for compressed images. The system is capable of authenticating images published in a compressed format by tolerating distortion introduced due to both embedding and compression while detecting other manipulations. Towards this end, a content-based fragile signature is derived and embedded into a compressed image using a robust watermarking technique. The scheme incorporates a mechanism to overcome the alteration of the fragile signature caused by the embedding and compression distortions. We build a general framework for the authentication based on the new concept of compensated signature embedding, such that media samples are iteratively adjusted to compensate for the distortion introduced and to restore the signature. The scheme was then customized for image data compressed using the JPEG2000 standard and tested after subjecting the compressed image to added degradations. Simulation results are presented to show that the proposed scheme successfully performs the authentication. The scheme does not require any third party reference or side information. The signature and embedding are designed to reveal tamper localization. The scheme can be modified to work with other types of compression methods and embedding schemes.