Side-informed steganography with additive distortion

Abstract

Side-informed steganography is a term used for embedding secret messages while utilizing a higher quality form of the cover object called the precover. The embedding algorithm typically makes use of the quantization errors available when converting the precover to a lower quality cover object. Virtually all previously proposed side-informed steganographic schemes were limited to the case when the side-information is in the form of an uncompressed image and the embedding uses the unquantized DCT coefficients to improve the security when JPEG compressing the precover. Inspired by the side-informed (SI) UNIWARD embedding scheme, in this paper we describe a general principle for incorporating the sideinformation in any steganographic scheme designed to minimize embedding distortion. Further improvement in security is obtained by allowing a ternary embedding operation instead of binary and computing the costs from the unquantized cover. The usefulness of the proposed embedding paradigm is demonstrated on a wide spectrum of various information-reducing image processing operations, including image downsampling, color depth reduction, and filtering. Side-information appears to improve empirical security of existing embedding schemes by a rather large margin.

Large-Scale JPEG Image Steganalysis Using Hybrid Deep-Learning Framework

“…We follow this approach in our research. Therefore, a grayscale input image can be represented as X = (x pq ) M×N = C + N, where C = (c pq ) M×N , c pq ∈ R denotes the corresponding cover image, and N = (n pq ) M×N , n pq ∈ R denotes the additive stego noise 10 .…”

“…That is to say, for each given z (2) rs , it is only the weighted sum of lower-layer inputs located in a m × n local area with index (r, s) as its centre irrespective of boundary condition, and the weights used in the weighted sum are shared in the calculation of all the z (2) rs , 1 ≤ r ≤ M, 1 ≤ s ≤ N. By rewriting (1) using two-dimensional indexing, setting l = 1, a (1) pq = x pq and restrict the size of the dot product to m × n (m and n assume to be odd to omit unimportant details), we get: (10) In (10), · denotes the ceiling operation. From (10) we can see that if the convolutional layer is initialized with kernels which are already sensitive to the stego noise (e.g. KV kernel) or is regularized as high-pass as proposed in [34], then m can be suppressed.…”

“…Therefore in either scenario, on average (10). 10 For JPEG steganography, the additive stego noise is directly added to quantized DCT coefficients. However, the linearity property of the DCT/IDCT transform guarantees that the corresponding stego noise in the spatial-domain representation is still additive.…”

“…However, analogous schemes have not yet been proposed in the JPEG domain. Although utilizing side information of a pre-cover image (raw or uncompressed) can improve the security of JPEG steganography [6], [7], [10], its applicability remains limited due to scarce availability of pre-cover images.…”

“…rs is proportional/inverse proportional to, or quasi-proportional/inverse quasi-proportional to z (2) rs provided the polarity of z (2) rs remains the same. Return to (10)…”