Fuzzy Authentication Algorithm with Applications to Error Localization and Correction of Images

Images are normally protected using standard messages authentication codes to protect them against tampering and forgeries. One problem with this approach is that when such images are transmitted over a noisy medium, even a single bit error might render the image as un-authentic to the receiver. In this paper, a noise tolerant data authentication algorithm is proposed. The proposed algorithm can perform authentication in the presence of minor errors but at the same time identify forgeries in the data. This algorithm is then extended by demonstrating its applications in image authentication. The extended algorithm is called as fuzzy authentication algorithm. It has the ability to localize errors in an image as well as correct the localized errors using error correcting codes. The proposed algorithm rejects only the potentially erroneous / unauthentic parts of the image and correct or authenticate the remaining parts if the number of errors is below a certain threshold. This reduces the need for retransmission of the complete image and only a few parts might be retransmitted if the application demands. This property is especially useful in real-time communications. It is better to obtain a part of the authentic image, rather than having no image at all. A security analysis of the proposed algorithm is given, and simulation results are presented to demonstrate its error localization and correction capabilities.