CSIS: Compressed sensing‐based enhanced‐embedding capacity image steganography scheme

Agrawal, Rohit; Ahuja, Kapil

doi:10.1049/ipr2.12161

Cited by 11 publications

(12 citation statements)

References 51 publications

(153 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…end for 10: end for 11: return t ′ î from subsection 2.1 that the size of the DCT sparsified vectors is (p 1 + p 2 ) × 1 with p 1 + p 2 = b 2 (here, b 2 = 64). Applying DCT on images results in a sparse vector where more than half of the coefficients have values that are either very small or zero (Agrawal and Ahuja, 2021;Pal et al, 2019). This is the case here as well.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

SABMIS: Sparse approximation based blind multi-image steganography scheme

Agrawal¹,

Ahuja²,

Steinbach³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Steganography is a technique of hiding secret data in some unsuspected cover media so that it is visually imperceptible. The secret data as well as the cover media may be text or multimedia. Image steganography, where the cover media is an image, is one of the most commonly used schemes. Here, we focus on image steganography where the hidden data is also an image. Specifically, we embed grayscale secret images into a grayscale cover image, which is considered to be a challenging problem. Our goal is to develop a steganography scheme with enhanced embedding capacity while preserving the visual quality of the stego-image and ensuring that the stego-image is resistant to steganographic attacks. Our proposed scheme involves use of sparse approximation and our novel embedding rule, which helps to increase the embedding capacity and adds a layer of security. The stego-image is constructed by using the Alternating Direction Method of Multipliers (ADMM) to solve the Least Absolute Shrinkage and Selection Operator (LASSO) formulation of the underlying minimization problem. This method has a fast convergence, is easy to implement, and also is extensively used in image processing. Finally, the secret images are extracted from the constructed stego-image using the reverse of our embedding rule. Using these components together helps us to embed up to four secret images into one cover image (instead of the common embedding of two secret images) and forms our most novel contribution. We term our scheme SABMIS (Sparse Approximation Blind Multi-Image Steganography). We perform extensive experiments on several standard images, and evaluate the embedding capacity, Peak Signal-to-Noise Ratio (PSNR) value, mean Structural Similarity (MSSIM) index, Normalized Cross-Correlation (NCC) coefficients, entropy, and Normalized Absolute Error (NAE). We obtain embedding capacities of 2 bpp (bits per pixel), 4 bpp, 6 bpp, and 8 bpp while embedding one, two, three, and four secret images, respectively. These embedding capacities are higher than all the embedding capacities obtained in the literature until now. Further, there is very little deterioration in the quality of the stego-image as compared to its corresponding cover image (measured by above metrics). The quality of the original secret images and their corresponding extracted secret images is also almost the same. Further, due to our algorithmic design, our scheme is resistant to steganographic attacks as well.

show abstract

Section: Resultsmentioning

confidence: 99%

“…For ADMM, we set the maximum number of iterations as 500, the absolute stopping tolerance as 1 × 10 −4 , and the relative stopping tolerance as 1 × 10 −2 . These values are again taken based upon our experience with a similar algorithm (Agrawal and Ahuja, 2021). Eventually, our ADMM almost always converges in 10 to 15 iterations.…”

Section: Resultsmentioning

confidence: 99%

SABMIS: Sparse approximation based blind multi-image steganography scheme

Agrawal¹,

Ahuja²,

Steinbach³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…In order to improve the persuasiveness of the proposed method, we compare HCISNet with the state-of-the-are methods and similar works, such as HiDDeN [20], SteganoGAN [23], CSIS [44], SteGAN [45] and HidingGAN [51]. We reproduce the HiDDeN and SteganoGAN, where the training batch sizes are set to six.…”

Section: Settingsmentioning

confidence: 99%

“…Besides, the other parameters are same as papers. However, since there are uncertain parameters in CSIS, SteGAN and HidingGAN, we refer to the results of papers [44,45,51].…”

Section: Settingsmentioning

confidence: 99%

“…Tang et al [43] propose an adaptive fuzzy inference approach for image steganography, and least significant bit substitution is used to hide the data adaptively. Agrawal et al [44] propose a Compressed Sensing Image Steganography (CSIS) scheme. It proposes a novel data extraction method to recover secret data lossless and achieves resistant to steganalysis attacks while embedding binary data in images.…”

Section: Steganographymentioning

confidence: 99%

See 1 more Smart Citation

HCISNet: Higher‐capacity invisible image steganographic network

Liu

et al. 2021

IET Image Processing

View full text Add to dashboard Cite

Information hiding as a crucial method for multimedia security and privacy protection, has received a substantial amount of attention. However, the most existing methods focus on the resistance of steganalysis and the robustness of information, resulting in low capacity. In this paper, to further increase the capacity of information hiding, a high-capacity adversarial image steganography model with end-to-end manner, termed as HCISNet is proposed. First, an enhanced Dense Atrous Spatial Pyramid Pooling module is presented to learn more semantic information in cover image, which can adaptively embed more message in redundant regions with rich textures. Then, a modified discriminator network with lightweight residual block is employed to assist the encoder network with optimal highcapacity solution. Meanwhile, the multiple objective function with the perceptual loss and several training tricks are developed, which can improve the visual and perceptual consistency of the original and the steganographic image, further enhancing the capacity. Finally, experiments and analysis are conducted on three public datasets, where this model has better imperceptibility, higher security and greater capacity. Under same experimental conditions, the cover image can embed 5.68 bits per pixel (BPP), far exceeding the previous highest value of 4.4 BPP in state-of-the-art methods in the authors' knowledge.This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

show abstract