Universal Deep Network for Steganalysis of Color Image Based on Channel Representation

Wei, Kangkang; Luo, Weiqi; Tan, Shunquan; Huang, Jiwu

doi:10.1109/tifs.2022.3196265

Cited by 16 publications

(5 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The user's requirement for image modification can be satisfied by image editing models (Zhu et al 2016;Wonka 2019, 2020;Zhu et al 2020;Choi et al 2021;Kim, Kwon, and Ye 2022;Patashnik et al 2021;Rombach et al 2022;Brooks, Holynski, and Efros 2022;Wei et al 2023) and controllable image generation methods (Meng et al 2022b;Park et al 2019;Nichol et al 2022;Hertz et al 2022;Couairon et al 2023;Orgad, Kawar, and Belinkov 2023;Parmar et al 2023;Feng et al 2023).…”

Section: Image Editing Methodsmentioning

confidence: 99%

Accelerating Text-to-Image Editing via Cache-Enabled Sparse Diffusion Inference

Yu,

Li,

et al. 2024

AAAI

View full text Add to dashboard Cite

Due to the recent success of diffusion models, text-to-image generation is becoming increasingly popular and achieves a wide range of applications. Among them, text-to-image editing, or continuous text-to-image generation, attracts lots of attention and can potentially improve the quality of generated images. It's common to see that users may want to slightly edit the generated image by making minor modifications to their input textual descriptions for several rounds of diffusion inference. However, such an image editing process suffers from the low inference efficiency of many existing diffusion models even using GPU accelerators. To solve this problem, we introduce Fast Image Semantically Edit (FISEdit), a cached-enabled sparse diffusion model inference engine for efficient text-to-image editing. The key intuition behind our approach is to utilize the semantic mapping between the minor modifications on the input text and the affected regions on the output image. For each text editing step, FISEdit can 1) automatically identify the affected image regions and 2) utilize the cached unchanged regions' feature map to accelerate the inference process. For the former, we measure the differences between cached and ad hoc feature maps given the modified textual description, extract the region with significant differences, and capture the affected region by masks. For the latter, we develop an efficient sparse diffusion inference engine that only computes the feature maps for the affected region while reusing the cached statistics for the rest of the image. Finally, extensive empirical results show that FISEdit can be 3.4 times and 4.4 times faster than existing methods on NVIDIA TITAN RTX and A100 GPUs respectively, and even generates more satisfactory images.

show abstract

Section: Image Editing Methodsmentioning

confidence: 99%

Accelerating Text-to-Image Editing via Cache-Enabled Sparse Diffusion Inference

Yu,

Li,

et al. 2024

AAAI

View full text Add to dashboard Cite

show abstract

“…PNet and VNet learned the a priori knowledge of DCTR and other frequency domain steganalysis and added a JPEG phase-aware module in the network framework to learn the signal-to-noise ratio information in the frequency domain. The parameters in the preprocessing layer were updated along with the network backpropagation during the training process of the full-learning network [33], [54], [55].The SR-Nert utilized the residual network to simulate the process of traditional SRM in filtering features, which could be applied not only in the spatial domain but also had good results in the JPEG domain. Vit exploited a convolutional visual transformer to capture local and global dependencies between noisy features for spatial domain information steganography.…”

Section: Related Workmentioning

confidence: 99%

“…2) Analysis of JEPG domain comparison experiment results: For the JPEG domain, PNet [53], VNet [53], DCTR [51], and UCNet [55] for payloads 0.1-0.5bpnzac(bit per non-zero AC DCT coefficient) are tested for quality factors 75. The results of the JEPG domain comparison experiment are listed in Tab.IV and Tab.V.…”

Section: B Comparison Experiments 1) Analysis Of Spatial Domain Compa...mentioning

confidence: 99%

RED-Net: Residual and Enhanced Discriminative Network for Image Steganalysis in the Internet of Medical Things and Telemedicine

Chen,

Zhou,

et al. 2024

IEEE J. Biomed. Health Inform.

View full text Add to dashboard Cite

Internet of Medical Things(IoMT) and telemedicine technologies utilize computers, communications, and medical devices to facilitate off-site exchanges between specialists and patients, specialists, and medical staff. If the information communicated in IoMT is illegally steganography, tampered or leaked during transmission and storage, it will directly impact patient privacy or the consultation results with possible serious medical incidents. Steganalysis is of great significance for the identification of medical images transmitted illegally in IoMT and telemedicine. In this paper, we propose a Residual and Enhanced Discriminative Network(RED-Net) for image steganalysis in the internet of medical things and telemedicine. RED-Net consists of a steganographic information enhancement module, a deep residual network, and steganographic information discriminative mechanism. Specifically, a steganographic information enhancement module is adopted by the RED-Net to boost the illegal steganographic signal in texturally complex high-dimensional medical image features. A deep residual network is utilized for steganographic feature extraction and compression. A steganographic information discriminative mechanism is employed by the deep residual network to enable it to recalibrate the steganographic features and drop highfrequency features that are mistaken for steganographic information. Experiments conducted on public and private datasets with data hiding payloads ranging from 0.1bpp/bpnzac-0.5bpp/bpnzac in the spatial and JEPG domain led to RED-Net's steganalysis error P E in the range of 0.0732-0.0010 and 0.231-0.026, respectively. In general, qualitative and quantitative results on public and private datasets demonstrate that the RED-Net outperforms 8 state-of-art steganography detectors.

show abstract

“…In order to make a more comprehensive comparison, three traditional manual steganalysis algorithms were implemented: maxSRMd2 [49] , PSRM [50] and DCTR [51] and five deep-learning-based steganalysis networks, PNet [29] , VNet [29] , SRNet [31] , UCNet [32] , and Vit [33] . All of them have been trained using the same datasets and have converged.…”

Section: Zhengyuan Zhou and Kai Chenmentioning

confidence: 99%

“…In 2019, Boroumand et al [31] proposed a deep residual network called SRNet, which is able to minimize the use of heuristics and externally enforced elements and achieve high detection accuracy among a total of five steganographic algorithms on spatial and JEPG domains. In 2022, Wei et al [32] designed a steganalysis method named UCNet that can be used for color images. This one first divides the input image into three channels and next extracts image residuals using 62 high-pass filters that are collected for analysis.…”

Section: Introductionmentioning

confidence: 99%

Global texture sensitive convolutional transformer for medical image steganalysis

Zhou,

Chen,

et al. 2024

Multimedia Systems

View full text Add to dashboard Cite

The use of medical images by hackers or illegal organizations as a vehicle for information leakage refers to steganography. Exchanged between PACS or communicated during telemedicine sessions, images are modified to hide data. Such leaks through stego-images may result in the disclosure of doctors' or patients' personal data, or of sensitive hospital data posing thus major risks in terms of privacy and security of the information system. In this paper, to detect these illegal image-based communications, we propose a steganalysis approach, the originality of which relies on a novel neural network GTSCT-Net. This one first extracts texture features as global texture features based on location specificity of different parts in image and then extracts possible steganographic information by composing mutlihead self-attention and deep convolution blocks. It also offers easier convergence and higher accuracy on a lower information embedding rate. Comparative experiments on private and public datasets show that the performance of GTSCT-Net for medical image intrusion detection is separately up to 10.12% and 2.97% better than recent advanced steganography detectors.

show abstract

Universal Deep Network for Steganalysis of Color Image Based on Channel Representation

Cited by 16 publications

References 49 publications

Accelerating Text-to-Image Editing via Cache-Enabled Sparse Diffusion Inference

Accelerating Text-to-Image Editing via Cache-Enabled Sparse Diffusion Inference

RED-Net: Residual and Enhanced Discriminative Network for Image Steganalysis in the Internet of Medical Things and Telemedicine

Global texture sensitive convolutional transformer for medical image steganalysis

Contact Info

Product

Resources

About