Multimedia authentication techniques are used to verify the information integrity, the alleged source of data, and the reality of data. Authentication of data integrity involves detecting manipulations to the multimedia data, and determining whether the changes are acceptable. Some authentication techniques consider the whole piece of multimedia data and do not allow for any manipulation. However, digital multimedia files are often compressed, which can complicate the authentication process. This technology is a semi-fragile watermarking technique that accepts JPEG lossy compression on the watermarked image by comparing transformed-domain image data, typically discrete cosine transform (DCT) coefficients. The algorithms can identify the positions of corrupted blocks, and recover them with approximation of the original ones. In addition to compression, adjustments of the brightness of the image within reasonable ranges are also acceptable using this authenticator.
This technology is able to authenticate images that have been altered using deterministic algorithms that compare DCT coefficients between transformed-domain image data. The authenticator is based on two invariant properties of DCT coefficients before and after JPEG compressions. Conventional techniques used to authenticate compressed image data rely on moment-based digital signatures or edge-based methods, which fail to detect malicious manipulations such as cropping and replacement, or exhibit excessive sensitivity to color alterations, respectively. This authentication technology is able to differentiate between malicious manipulations and acceptable, legitimate alterations such as compression, filtering, or transformation of encoding format. In addition to authentication, the technology can also recover approximate pixel values in corrupted areas by embedding recovery bits into the watermarked image.
The technology was tested on a photoshop compressed and manipulated 256x256 gray-level image in order to determine image degradation due to watermarking, probability of false alarm, probability of miss and probability of successful attack.
Patent Issued (US 6,879,703)
Tech Ventures Reference: IR M01-018