Performance analysis of Fridrich-Goljan self-embedding authentication method

This paper analyzes the performance of the image authentication method based on robust hashing proposed by J. Fridrich and M. Goljan. In this method both the embedder and the detector generate the watermark from a perceptual digest of the image. Therefore, an accurate performance analysis requires assessing the relation between noise and hash bit errors. Our approach first derives the probability of hash bit error due to watermark embedding and/or the attack, and then uses such probability to derive the probabilities of false positive and false negative. Index Terms Content-based authentication, robust hash, performance analysis.