Abstract: A feature based fragile image watermarking framework is proposed to resist vector quantization (VQ) attack. Image features are extracted to process the original watermark or generate the watermark so that the embedded watermark is not only unknown to the attacker, but also dependent on the image. Therefore, different images have different embedded watermarks so that it is impossible for the attacker to build VQ codebook. At the same time, the watermark is embedded into the image under consideration of security and localization．The basic requirements for the embedding scheme are given for the trade-off between the security and localization. An actual moment invariants based fragile image watermarking scheme IS presented to demonstrate our fragile watermarking framework. Analysis and experimental results demonstrate that our algorithm can simultaneously detect even one bit image alteration with graceful localization and resist vector quantization attack without the need of any unique keys or image indexes.