Abstract: A novel self-embedding fragile watermarking technique is proposed for image integrity authentication, where image-dependent uncertainty is introduced to overcome the security problem existing in most algorithms. This purpose is implemented by using an image-dependent key that is obtained by combining image features with one camera key. The problem of key management is also solved because of the uniqueness of camera key. In watermark embedding, image message is used as the fragile watermark so that no reference watermark is needed in authentication. Performance analysis is given, which shows that larger probability of tamper detection can be achieved when larger amount of image message is embedded. Possible attacks are also analyzed and the probabilities of successful attack and detection are given respectively. Finally, the experimental results are presented to demonstrate its effectiveness.