This study examines the feasibility of using statistics to perform embedded signature-based authentication. Doing so is an attempt to view a reference watermark embedded with signature as side information. The reference watermark is extracted to determine channel statistics, based on the Bayes theorem, and used to extract the embedded signature. The reliability of the extracted signature and the uncertainty of the channel status are also measured. Experimental results demonstrate that the reliability and uncertainty measures based on channel statistics are meaningful, and that the embedded signature can survive high-quality JPEG compression and manipulation such as negation.
Relation:
Proceedings of 5th IEEE International Workshop on Multimedia Signal Processing, US Virgin Islands, U. S. A
