The security flaws in some authentication watermarking schemes

Watermarking technology was originally proposed for copyright protection. Recently it has been applied to media authentication so that a proof of authenticity is inserted into the media instead of being appended to the media as a separated attachment. However, security requirements of the authentication are overlooked in some authentication watermark schemes. In this paper we analyze three authentication watermarking schemes and point out their security flaws. The first scheme is the color authentication scheme in [S.C. Byun et al., 2002]. The scheme is not secure in the sense that as long as an attacker obtains one authenticated image, he is able to forge authentic images without the secret key. The second scheme [Ping Wah Wing, et al., 2001] is an authentication scheme but it is extended for ownership incorrectly. The third one, the robust invertible watermarking scheme [J. Friedrich et al., 2002], employs a multiple of secret random sequences to produce a watermark. However these sequences are independent of the original images, i.e., they remain invariable for different images. An adversary, having sufficient number of original images, can reconstruct the secret sequences by solving simultaneous equations. With these reconstructed sequences, the attacker can forge authentic image freely. The attack can be thwarted with content related sequences generated from both the secret key and the original image.