Image Steganography by Deep CNN Auto-Encoder Networks

Kich, Ismail

doi:10.30534/ijatcse/2020/75942020

Cited by 7 publications

(9 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the first experiments, we trained our model on images with different sizes; the images were randomly selected and then resized from the ImageNet dataset. The results were then compared with our previous method cited in [7]. To accomplish the comparison, we first test the proposed method using the same loss function used in the previous model [7]; we call it "Proposed L 2 +L 2 ".…”

Section: Experiments Resultsmentioning

confidence: 99%

“…The results were then compared with our previous method cited in [7]. To accomplish the comparison, we first test the proposed method using the same loss function used in the previous model [7]; we call it "Proposed L 2 +L 2 ". Then, we use the loss function proposed in this article (see Section 3.4); we call it "Proposed L 1 + V + L 2 ".…”

Section: Experiments Resultsmentioning

confidence: 99%

“…Third, a new loss function is used during training, so that the generated stego image, which is controlled by steganalysts, can preserve high fi-delity. Our previous work [7] was built with a basic convolution layers and trained with loss function based on weighted L 2 loss. The test results suggest that the proposed model outperforms our previous model and existing models.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

CNN Auto-Encoder Network Using Dilated Inception for Image Steganography

Kich¹,

Ameur²,

Taouil³

2021

IJFIS

Self Cite

View full text Add to dashboard Cite

Numerous studies have used convolutional neural networks (CNNs) in the field of information concealment as well as steganalysis, achieving promising results in terms of capacity and invisibility. In this study, we propose a CNN-based steganographic model to hide a color image within another color image. The proposed model consists of two sub-networks: the hiding network is used by the sender to conceal the secret image; and the reveal network is used by the recipient to extract the secret image from the stego image. The architecture of the concealment sub-network is inspired by the U-Net auto-encoder and benefits from the advantages of the dilated convolution. The reveal sub-network is inspired by the auto-encoder architecture. To ensure the integrity of the hidden secret image, the model is trained end to end: rather than training separately, the two sub-networks are trained simultaneously a pair of networks. The loss function is elaborated in such a way that it favors the quality of the stego image over the secret image as the stego image is the one that comes under steganalysis attacks. To validate the proposed model, we carried out several tests on a range of challenging publicly available image datasets such as ImageNet, Labeled Faces in the Wild (LFW), and PASCAL-VOC12. Our results show that the proposed method can dissimulate an image into another one with the same size, reaching an embedding capacity of 24 bit per pixel without generating visual or structural artefacts on the host image. In addition, the proposed model is generic, that is, it does not depend on the image's size or the database source.

show abstract

Section: Experiments Resultsmentioning

confidence: 99%

Section: Experiments Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

CNN Auto-Encoder Network Using Dilated Inception for Image Steganography

Kich¹,

Ameur²,

Taouil³

2021

IJFIS

Self Cite

View full text Add to dashboard Cite

show abstract

“…The image patches are cropped to 224 throughout the learning process. A simple and effective learning rate of 0.001 is used from formula [17], [19]- [20], with drop factor of 0.3. mini batch size of 8 is employed, this is maintained for both the upsampling and downsampling respectively. Our architecture is optimized using stochastic gradient descent with momentum of technique of 0.9, where best hyper-parameters are selected.…”

Section: Corollarymentioning

confidence: 99%

“…Comparison of our proposed model with conventional method on the Food-101 dataset[20]. The best results are bolded.…”

mentioning

confidence: 99%

Semantic Parsing for Automatic Retail Food Image Recognition

2020

IJATCSE

View full text Add to dashboard Cite

Explosion of the world of computer vision, pave the way to visual recognition which is now extended to visual food and metadata recognition. Physical activities disruption due to Covid-19 pandemic rapidly increases the online food order. Online Customers relies on available good quality food images and metadata information, to make decision of the type of food to order. Recently developed deep learning networks outperform the classical approaches for visual food recognition, however these networks thirsts for large datasets to achieve best performance. Majority population of present online customers order food to small-scale restaurants, these restaurants deals with small datasets, thus restricted them to take advantage of modern tools and to participate in the billion-digital business. In this work, we modified the existing deep-CNN architecture to fit the small-scale restaurant dataset and trained on an end-to-end DeepLab v3+ initialized from ResNet weight. Our proposed novel architecture is designed by exploiting the output of multiscale contextual information of CNN encoder and fed in the low-level features of our constructed Resnet-18 as the backbone network, and finally fine-tuned with simple filters, and bilinear interpolation on the order factor by 4. This approach reduces the serious overfitting of the deep-CNN. The metadata recognition was done using enhanced-OCR, where the segmented image was analyzed at high-level layers. The accuracy of our method is reported using IoU and BF score. The numerical validation of the method is carried out on ETH-food-101 dataset and it demonstrates compliance to the state-of-the-art performance.

show abstract

A Comparative Analysis of Various Loss Functions for Image Data Hiding Using Deep Neural Network

Chanchal

Malathi

2022

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

Image Steganography by Deep CNN Auto-Encoder Networks

Cited by 7 publications

References 25 publications

CNN Auto-Encoder Network Using Dilated Inception for Image Steganography

CNN Auto-Encoder Network Using Dilated Inception for Image Steganography

Semantic Parsing for Automatic Retail Food Image Recognition

A Comparative Analysis of Various Loss Functions for Image Data Hiding Using Deep Neural Network

Contact Info

Product

Resources

About