IMAGE STEGANOGRAPHY DETECTION USING RESNET AND SRNET MODELS

DOI: 10.31673/2412-4338.2026.019009

Authors

Abstract

This paper presents a comprehensive analysis of the effectiveness of modern deep convolutional neural network architectures—SRNet and ResNetV2 (ResNet50V2, ResNet101V2, ResNet152V2)—in spatial digital image steganalysis tasks. Primary attention is devoted to investigating the role of the high-pass filtering preprocessing block (HPF-filter block) and the influence of the filter count and filtering block architecture on the detection accuracy of weak steganographic signals introduced by the LSB method.

The CIFAR-10 and LabelMe datasets were utilized to form training and testing samples, serving as the basis for creating artificial sets of cover/stego images using LSB steganography with UTF-8 encoding support and controlled payload capacity.

It was established that the canonical SRNet architecture requires the integration of an additional HPF block to achieve stable convergence within a small number of training epochs, demonstrating high sensitivity to anomalies but significant RAM consumption, which limits the training dataset size in cloud environments. In contrast, ResNetV2-based models proved to be more scalable and practical for real-time monitoring systems, although their efficiency critically depends on the configuration of a multi-scale filter block (3×3, 5×5, 7×7).

A directional multi-channel HPF module approach with orientational selectivity is proposed, providing an optimal balance between detection accuracy and computational overhead. Permitting the training of high-pass filters provides an additional increase in accuracy, confirming the feasibility of a hybrid approach that combines a priori knowledge of digital signal processing with deep learning. The results obtained can be utilized in the design of practical digital content monitoring systems and allow for the formulation of recommendations for selecting AI models depending on the application scenario: from precision research tools to industrial information security systems.

Keywords: steganalysis, deep learning, convolutional neural networks, ResNetV2, SRNet, high-pass filtering, LSB steganography.

References

  1. Fridrich J. Steganography in Digital Media: Principles, Algorithms, and Applications. Cambridge University Press, 2009. 437 p.
  2. Xu G., Wu H.Z., Shi Y.Q. Structural design of convolutional neural networks for steganalysis. IEEE Signal Processing Letters, 2016, vol. 23, no. 5, pp. 708-712. DOI: 10.1109/LSP.2016.2548421
  3. Fridrich J., Kodovsky J. Rich models for steganalysis of digital images. IEEE Transactions on Information Forensics and Security, 2012, vol. 7, no. 3, pp. 868-882. DOI: 10.1109/TIFS.2012.2190402
  4. He K., Zhang X., Ren S., Sun J. Deep residual learning for image recognition. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770-778. DOI: 10.1109/CVPR.2016.90
  5. Xu G. Deep convolutional neural network to detect J-UNIWARD. In: Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security, 2017, pp. 67-73. DOI: 10.1145/3082031.3083236
  6. Ye J., Ni J., Yi Y. Deep learning hierarchical representations for image steganalysis. IEEE Transactions on Information Forensics and Security, 2017, vol. 12, no. 11, pp. 2545-2557. DOI: 10.1109/TIFS.2017.2710946
  7. Zhang R., Zhu F., Liu J., Liu G. Depth-wise separable convolutions and multi-level pooling for an efficient spatial CNN-based steganalysis. IEEE Transactions on Information Forensics and Security, 2020, vol. 15, pp. 1138-1150. DOI: 10.1109/TIFS.2019.2936913
  8. Boroumand M., Chen M., Fridrich J. Deep residual network for steganalysis of digital images. IEEE Transactions on Information Forensics and Security, 2019, vol. 14, no. 5, pp. 1181-1193. DOI: 10.1109/TIFS.2018.2871749
  9. Tabares-Soto R., Arteaga-Arteaga H.B., Buritica O.M.A., et al. Deep learning for steganalysis: evaluating model robustness against image transformations. Frontiers in Artificial Intelligence, 2025, vol. 8, article 1532895. DOI: 10.3389/frai.2025.1532895
  10. Veit A., Wilber M.J., Belongie S. Residual networks behave like ensembles of relatively shallow networks. In: Advances in Neural Information Processing Systems (NeurIPS), 2016, vol. 29, pp. 550-558.
  11. Ntivuguruzwa J.-P., Kurundayev M., Ullah M., et al. A convolutional neural network to detect possible hidden data in spatial domain images. Cybersecurity, 2023, vol. 6, article 32. DOI: 10.1186/s42400-023-00156-x
  12. Ozcan S., Mustacoglu A.F. Transfer learning effects on image steganalysis with pre-trained deep residual neural network model. In: IEEE International Conference on Big Data, 2018, pp. 2280-2287. DOI: 10.1109/BigData.2018.8622437
  13. Yedroudj M., Comby F., Chaumont M. Yedroudj-Net: An efficient CNN for spatial steganalysis. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018, pp. 2092-2096. DOI: 10.1109/ICASSP.2018.8461438
  14. Dwaik A., Nandi A.K., Naous T., Alani S., Alsarhan A. Enhancing the performance of convolutional neural network image-based steganalysis in spatial domain using Spatial Rich Model and 2D Gabor filters. Journal of Information Security and Applications, 2024, vol. 85, article 103862. DOI: 10.1016/j.jisa.2024.103862
  15. Wang Z., Gao N., Wang X., Qu X., Li L., Zhang X. Deep learning for steganalysis of diverse data types: A review of methods, taxonomy, challenges and future directions. arXiv preprint arXiv:2308.04522, 2024.
  16. Tabares-Soto R., Arteaga-Arteaga H.B., Orozco-Arias S., et al. Strategy to improve the accuracy of convolutional neural network architectures applied to digital image steganalysis in the spatial domain. PeerJ Computer Science, 2021, vol. 7, e451. DOI: 10.7717/peerj-cs.451
  17. Liu F., Zhou X., Yan X., Peng J., Hu Y., Chen Q. Preprocessing enhancement method for spatial domain steganalysis. Mathematics, 2022, vol. 10, no. 21, article 3936. DOI: 10.3390/math10213936
  18. Wang X., Liao D., Dai Y., Li H. A steganalysis framework based on CNN using the filter subset selection method. Multimedia Tools and Applications, 2020, vol. 79, pp. 21307-21326. DOI: 10.1007/s11042-020-08831-8
  19. Jin Z., Yang Y., Chen Y., Chen Y. IAS-CNN: Image adaptive steganalysis via convolutional neural network combined with selection channel. Journal of Sensors, 2020, article 1550147720911002. DOI: 10.1177/1550147720911002
  20. Chaumont, Marc. (2020). Deep learning in steganography and steganalysis. 10.1016/B978-0-12-819438-6.00022-0. In book: Digital Media Steganography (pp.321-349)
  21. Krizhevsky, A. (2009). Learning Multiple Layers of Features from Tiny Images. https://www.cs.toronto.edu/~kriz/learning-features-2009-TR.pdf.
  22. Stefanek G., Gulbransen L., Spink G., Morawski J., Filla D., Rabello De Castro R. A comparison of ai models to detect hidden messages in images. (2024). Issues In Information Systems. 119-132.  DOI: 10.48009/3_iis_2024_110
  23. The LabelMe-12-50k dataset. URL: https://www.ais.uni-bonn.de/download/datasets.html
  24. Meike Helena Kombrink, Zeno Jean Marius Hubert Geradts, and Marcel Worring. 2024. Image Steganography Approaches and Their Detection Strategies: A Survey. ACM Comput. Surv. 57, 2, Article 33 (February 2025), 40 pages. DOI: 10.1145/3694965
  25. S. Wu, S. -h. Zhong and Y. Liu, "A Novel Convolutional Neural Network for Image Steganalysis With Shared Normalization," in IEEE Transactions on Multimedia, vol. 22, no. 1, pp. 256-270, Jan. 2020, DOI: 10.1109/TMM.2019.2920605.
  26. Dwaik, A., & Belkhouche, Y. (2024). Enhancing the performance of convolutional neural network image-based steganalysis in spatial domain using Spatial Rich Model and 2D Gabor filters. Journal of Information Security and Applications, 85, 103864. DOI: 10.1016/j.jisa.2024.103864

Downloads

Published

2026-04-01

Issue

No. 1 (2026)

Section

Articles