A hybrid image authentication method robust to JPEG attacks

Checking authentication of digital images have become a necessity due to ever increasing use in many fields. Passive methods, which make use of image statistics, have gained popularity among researchers lately. Block based authentication methods using feature vectors of subblocks in images can be classified as either spatial or frequency domain. A new hybrid authentication technique is proposed in this paper with desired properties of both spatial and frequency domain methods in feature extraction. Both information from the color components and frequency domain representation of the corresponding blocks are used while forming the feature vectors. The proposed method is robust to some post processing operations such as Gauss blurring, additive white noise and JPEG compression as shown in the experimental results.