This paper presents a robust detection approach for the hidden information in the discrete cosine transform (DCT) domain.The proposed approach uses the α -stable distributions

which are more accurate probability models than the gaussian and generalized gaussian distributions for those heavy tailed non-gaussian data with the algebraic tails

to model AC DCT coefficients with distinct impulse distributional behavior.By applying the statistical inference on the probability models under the generalized Neyman-Pearson criterion

the robust detection structure is then developed.The resulting detector is based on the statistical signal detection theory and it exploits the imperceptibility constraints of the hidden information.Experiments show that:a) the novel detector is superior over the matched filter and b) though the performance of the novel detector is close to that of the detector based on the generalized gaussian models

the proposed detector does not need the "point elimination process"

which simplifies the realization of the new detector.