Iterative self-correction for secured images using turbo codes and soft input decryption.

An algorithm for improved error correction of the data protected using standard security mechanisms, like Message Authentication Codes, is introduced in this paper. Images are taken as example data, and the correction of images using the additionally available authentication data is investigated. Parts of the protected image, the so called "landmarks", are considered as the Region of Interest and protected by their respective authentication tags. An iterative process is used as a basis for the learning ability of the algorithm: in every iteration, parts of the decoder's Trellis path are learned by the algorithm, thereby improving the channel decoding results. In this way, the knowledge gained in the previous iterations of the algorithm is used in the current iteration for further improvement of decoding results. Additionally, iterative processes used for error corrections are supported by authentication tags, which are used as a measure of correction success. Simulation results for images are presented to show the effectiveness of the proposed algorithm. A security analysis of the proposed algorithm is also given in the paper.