Fragile watermarking scheme based on the block-wise dependence in the wavelet domain

In transform-domain fragile watermarking schemes for authentication purposes, a common practice is to watermark some selected transform coefficients in order to achieve low embedding distortion. However, we point out in this work that leaving most of the coefficients, usually the low frequency and zero-valued ones, unmarked opens wide security gap for attacks to be mounted on them. In this work, a fragile watermarking scheme is proposed to implicitly watermark all the coefficients by registering the zero-valued coefficients with a key-generated binary sequence so as to create the watermark and involving the unwatermarkable coefficients within a hierarchical neighbourhood of the watermarkable coefficients in a random manner during the embedding process. The benefit of this approach is twofold: Firstly, only a small amount of coefficients are marked while all the coefficients are protected, so that high transparency is achieved. Secondly, when the relationship established in a random manner within a hierarchical neighbourhood is disturbed, the watermark detector would fail to extract the watermarking. This puts up resistance to the vector quantisation attack and transplantation attack.