Abstract: Copy-paste with edge blurring manipulation is the commonly used tampering trick,and the generated forgery image is difficult to authenticate.A tampering image authentication algorithm based on dark channel is proposed,it focuses on detecting the forgery image generated by copy-paste with edge blurring manipulation.Dark channel is a novel image attribute.Because the number of nonzero pixels in blur region is higher than the one of intact regions,dark channel is exploitable for rapidly distinguishing the tampering region.The normalized image segmentation is applied for generating the semantic regions,these regions are described in Hilbert space,and their normalized correlation coefficients are used for determining the copy-paste source.Experiments demonstrate that the proposed approach can increase the average detection accuracy 6%,regardless of illumination variation,noise and JPEG compression ratio change.