Fragile watermarking using the VW2D watermark

Two classes of digital watermarks have been developed to protect the copyright ownership of digital images. Robust watermarks are designed to withstand attacks on an image (such as compression or scaling), while fragile watermarks are designed to detect minute changes in an image. Fragile marks can also identify where an image has been altered. This paper compares two fragile watermarks. The first method uses a hash function to obtain a digest of the image. An altered or forged version of the original image is then hashed and the digest is compared to the digest of the original image. If the image has changed the digests will be different. We will describe how images can be hashed so that any changes can be spatially localized. The second method uses the Variable-Watermark Two-Dimensional algorithm (VW2D) [1]. The sensitivity to changes is user-specific. Either no changes can be permitted (similar to a hard hash function), or an image can be altered and still be labeled authentic. Latter algorithms are known as semi-fragile watermarks. We will describe the performance of these two techniques and discuss under what circumstances one would use a particular technique.