Journals / Fırat Üniversitesi Mühendislik Bilimleri Dergisi / 2026 / Cilt: 38 Sayı: 2

Detection of LSB-Based Steganographic Data in Image Files Using Deep Learning Models

Resim Dosyalarında LSB Yöntemi ile Gizlenen Steganografik Verilerin Derin Öğrenme Modelleri Kullanarak Tespit Edilmesi

Emir Öztürk(TRAKYA ÜNİVERSİTESİ, MÜHENDİSLİK FAKÜLTESİ, BİLGİSAYAR MÜHENDİSLİĞİ BÖLÜMÜ)0000-0002-3734-5171
Andaç Mesut(Trakya Üniversitesi, Mühendislik ve Mimarlık Fakültesi, Bilgisayar Mühendisliği Bölümü, Edirne, Türkiye)0000-0003-0853-5256
Özlem Aydın Fidan(Trakya Üniversitesi, Mühendislik ve Mimarlık Fakültesi, Bilgisayar Mühendisliği Bölümü, Edirne, Türkiye)0000-0002-6401-4183
Pages
575–587
Publication Date
2026-09-30
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

The results showed that at low hiding ratios the performances of all models were close, while at high hiding ratios the distortions became more visible and the discrimination ability increased.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:In this study, the detectability of data hidden with an LSB-based steganography method was examined using deep learning models.
📊
Key Findings & Data:For the analysis, a stego dataset presented in the literature was used, and the results of experiments conducted at different hiding ratios (hr), bits per pixel values (hb), and image resolutions were evaluated using CNN-based models (ResNet-50, VGG, DenseNet, EfficientNet, MobileNet) and transformer-based models (ViT, DeiT, Swin, BEiT).
💡
Conclusion & Impact:The results showed that at low hiding ratios the performances of all models were close, while at high hiding ratios the distortions became more visible and the discrimination ability increased.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:CNN · steganografi · steganaliz

Abstract

In this study, the detectability of data hidden with an LSB-based steganography method was examined using deep learning models. For the analysis, a stego dataset presented in the literature was used, and the results of experiments conducted at different hiding ratios (hr), bits per pixel values (hb), and image resolutions were evaluated using CNN-based models (ResNet-50, VGG, DenseNet, EfficientNet, MobileNet) and transformer-based models (ViT, DeiT, Swin, BEiT). In addition, a hybrid model was proposed in order to compare the results, and its outcomes were discussed. The results showed that at low hiding ratios the performances of all models were close, while at high hiding ratios the distortions became more visible and the discrimination ability increased. Among the CNN architectures, ResNet-50 achieved the highest performance with an average F1 score of 0.88. In the transformer group, Swin Transformer stood out in all scenarios and reached the highest F1 score of 0.91. In contrast, BEiT showed lower performance than the other transformer architectures at both low and high hiding ratios and obtained an average F1 score of 0.75. The findings indicate that both transformer- and CNN-based approaches can be used effectively for steganalysis, while the general trend shows higher performance and representation capability for transformer architectures, although some CNN models produced comparable results.

Özet

Bu çalışmada, LSB tabanlı steganografi yöntemi ile gizlenmiş verilerin tespit edilebilirliği derin öğrenme modelleri kullanılarak incelenmiştir. İnceleme için literatürde sunulan stego veri seti kullanılmış ve Evrişimsel Sinir Ağı (CNN) tabanlı ResNet-50, VGG, DenseNet, EfficientNet, MobileNet ve transformer tabanlı ViT, DeiT, Swin, BEiT modelleri kullanılarak farklı gizleme oranları, piksel başına bit sayıları ve çözünürlüklerde gerçekleştirilen deneylerden elde edilen sonuçlar değerlendirilmiştir. Ayrıca sonuçların karşılaştırılması amacıyla bir hibrit model önerisi sunulmuş ve sonuçlar irdelenmiştir. Değerlendirme sonuçları incelendiğinde düşük gizleme oranlarında tüm modellerin performanslarının birbirine yakın olduğu, yüksek gizleme oranlarında ise bozulmaların daha belirgin hale gelmesiyle ayrım gücünün arttığı gözlenmiştir. CNN mimarileri içerisinde ResNet-50 ortalama 0,88 F1 skoru ile en yüksek başarıyı sergilerken, transformer grubunda Swin Transformer tüm senaryolarda öne çıkmış ve en yüksek F1 skoru olan 0,91 değerine ulaşmıştır. Buna karşılık BEiT modeli hem düşük hem de yüksek gizleme oranlarında diğer transformer mimarilerinden düşük performans göstermiştir ve ortalama 0,75 F1 skoru elde etmiştir. Elde edilen bulgular, steganaliz görevinde hem transformer hem de CNN tabanlı yaklaşımların etkili biçimde kullanılabildiğini ortaya koymaktadır. Bununla birlikte, genel eğilim transformer tabanlı mimarilerin CNN’lere kıyasla daha yüksek performans ve temsil gücü sunduğunu göstermektedir. Öte yandan, bazı CNN mimarilerinin de transformer mimarileri kadar başarılı sonuçlar ürettiği görülmüştür.

🔖

Cite This Article

Öztürk, E., Mesut, A., Fidan, Ö. A. (2026). Detection of LSB-Based Steganographic Data in Image Files Using Deep Learning Models. Fırat Üniversitesi Mühendislik Bilimleri Dergisi, Cilt: 38 Sayı: 2, 575–587. https://doi.org/10.35234/fumbd.1891591

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (48)

  1. Kumar N, Soni M. Randomized salt LSB: a novel approach to steganography with improved concealment and robustness. Int J Inf Technol 2025; 1-14.
  2. Putra ABW, Jufri MT, Lantara D, Assyauqi A, Wajiansyah A, Pranolo A. Steganography for data hiding in digital audio data using combined least significant bit and 4-wrap length method. In: 2018 2nd East Indonesia Conference on Computer and Information Technology (EIConCIT), 2018; IEEE. pp. 336-339.
  3. Sharma VK, Soni BP, Janu N, Aziz S, Shekhawat D. Review on image steganography using different LSB methods. J Discrete Math Sci Cryptogr 2024; 27(4): 1319-1329.
  4. Xiang L, Qin J, Xiang X, Tan Y. Text coverless information hiding based on BERT entity recognition. Hum-Centric Comput Inf Sci 2024; 14.
  5. Su W, Ni J, Sun Y. Stegastylegan: Towards generic and practical generative image steganography. In: Proceedings of the AAAI Conference on Artificial Intelligence, 2024. pp. 240-248.
  6. Sk A, Chakraborty S, Sen S. A novel blocked based U-Net model for image steganography. Multimed Tools Appl 2025; 1-20.
  7. Li N, Qin J, Xiang X, Tan Y. Robust coverless video steganography based on inter-frame keypoint matching. J Inf Secur Appl 2023; 79: 103653.
  8. Yang L, Xu D, Wang R, He S. Adaptive HEVC video steganography based on distortion compensation optimization. J Inf Secur Appl 2023; 73: 103442.
  9. Mou C, Xu Y, Song J, Zhao C, Ghanem B, Zhang J. Large-capacity and flexible video steganography via invertible neural network. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023. pp. 22606-22615.
  10. Meng L, Jiang X, Zhang Z, Li Z, Sun T. A robust coverless video steganography based on maximum DC coefficients against video attacks. Multimed Tools Appl 2024; 83(5): 13427-13461.
  11. Li J, Wang K, Jia X. A coverless audio steganography based on generative adversarial networks. Electronics 2023; 12(5): 1253.
  12. Xiang X, Tan Y, Qin J, Tan Y. Advancements and challenges in coverless image steganography: A survey. Signal Process 2025; 228: 109761.
  13. Neeta D, Snehal K, Jacobs D. Implementation of LSB steganography and its evaluation for various bits. In: 2006 1st International Conference on Digital Information Management, 2006; IEEE. pp. 173-178.
  14. Yang CH, Weng CY, Wang SJ, Sun HM. Adaptive data hiding in edge areas of images with spatial LSB domain systems. IEEE Trans Inf Forensics Secur 2008; 3(3): 488-497.
  15. Öztürk E, Mesut AŞ, Fidan ÖA. A character based steganography using masked language modeling. IEEE Access 2024; 12: 14248-14259.
  16. Mesut AŞ, Aydın Ö, Öztürk E. Anlamsal yöntemler kullanan bir metin steganografi uygulaması. In: Proceedings of the 1st International Symposium on Digital Forensics and Security (ISDFS), 2013. pp. 21-24.
  17. Aydın Ö, Mesut AŞ, Öztürk E. Finding the optimal color channel for information hiding in LSB insertion method. J Fundam Sci Appl 2020; 26(1): 1-5.
  18. Kich I, Ameur EB, Taouil Y, Benhfid A. Image steganography scheme using dilated convolutional network. In: 2021 12th International Conference on Information and Communication Systems (ICICS), 2021; IEEE. pp. 305-309.
  19. Xu Y, Mou C, Hu Y, Xie J, Zhang J. Robust invertible image steganography. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. pp. 7875-7884.
  20. Zhang C, Gao X, Liu X, Hou W, Yang G, Xue T, Wang L, Liu L. IDGAN: Information-driven generative adversarial network of coverless image steganography. Electronics 2023; 12(13): 2881.
  21. Kheddar H, Hemis M, Himeur Y, Megias D, Amira A. Deep learning for steganalysis of diverse data types: A review of methods, taxonomy, challenges and future directions. Neurocomputing 2024; 581: 127528.
  22. Chen Z, Huang L, Miao H, Yang W, Meng P. Steganalysis against substitution-based linguistic steganography based on context clusters. Comput Electr Eng 2011; 37(6): 1071-1081.
  23. Xu Q, Zhang R, Liu J. Linguistic steganalysis by enhancing and integrating local and global features. IEEE Signal Process Lett 2023; 30: 16-20.
  24. Zou J, Yang Z, Zhang S, ur Rehman S, Huang Y. High-performance linguistic steganalysis, capacity estimation and steganographic positioning. In: International Workshop on Digital Watermarking, 2020; Springer. pp. 80-93.
  25. Yang J, Yang Z, Ge X, Zou J, Gao Y, Huang Y. Link: Linguistic steganalysis framework with external knowledge. In: ICASSP 2023 - IEEE International Conference on Acoustics, Speech and Signal Processing, 2023; IEEE. pp. 1-5.
  26. Yang MB, Pang K, Wang H, Huang Y. Towards next-generation steganalysis: LLMs unleash the power of detecting steganography. arXiv preprint arXiv:2405.09090, 2024.
  27. Pevný T, Bas P, Fridrich J. Steganalysis by subtractive pixel adjacency matrix. In: Proceedings of the 11th ACM Workshop on Multimedia and Security, 2009. pp. 75-84.
  28. Fridrich J, Kodovsky J. Rich models for steganalysis of digital images. IEEE Trans Inf Forensics Secur 2012; 7(3): 868-882.
  29. Gao X, Yi J, Liu L, Tan L. A generic image steganography recognition scheme with big data matching and an improved ResNet50 deep learning network. Electronics 2025; 14(8): 1610.
  30. Liu P, Li S. Steganalysis of intra prediction mode and motion vector-based steganography by noise residual convolutional neural network. In: IOP Conference Series: Materials Science and Engineering, 2020; IOP Publishing. 012068.
  31. Yousfi Y, Butora J, Khvedchenya E, Fridrich J. ImageNet pre-trained CNNs for JPEG steganalysis. In: 2020 IEEE International Workshop on Information Forensics and Security (WIFS), 2020; IEEE. pp. 1-6.
  32. Yousfi Y, Butora J, Fridrich J, Fuji Tsang C. Improving EfficientNet for JPEG steganalysis. In: Proceedings of the 2021 ACM Workshop on Information Hiding and Multimedia Security, 2021. pp. 149-157.
  33. Hong E, Lim K, Oh TW, Jang H. Lightweight image steganalysis with block-wise pruning. Sci Rep 2023; 13(1): 16148.
  34. Ozcan S, Mustacoglu AF. Transfer learning effects on image steganalysis with pre-trained deep residual neural network model. In: 2018 IEEE International Conference on Big Data (Big Data), 2018; IEEE. pp. 2280-2287.
  35. Yang J, Shi YQ, Wong EK, Kang X. JPEG steganalysis based on DenseNet. arXiv preprint arXiv:1711.09335, 2017.
  36. Bravo-Ortiz MA, Mercado-Ruiz E, Villa-Pulgarin JP, Hormaza-Cardona CA, Quinones-Arredondo S, Arteaga-Arteaga HB, Orozco-Arias S, Cardona-Morales O, Tabares-Soto R. CVTStego-Net: A convolutional vision transformer architecture for spatial image steganalysis. J Inf Secur Appl 2024; 81: 103695.
  37. Delgado-Martínez MC, Ramirez Rodriguez FA, Guevara-Navarro E, Cardona-Pérez LM, Garcia-Herrera S, Isaza G, Tabares-Soto R, Bravo-Ortiz MA. R-SIT: A Swin-Transformer-based architecture for spatial image steganalysis. In: Iberoamerican Congress on Pattern Recognition (CIARP), November 2025; Cham: Springer Nature Switzerland. pp. 99-114.
  38. Wei KK, Luo WQ, Tan SQ, Huang JW. CTNet: A convolutional transformer network for color image steganalysis. J Comput Sci Technol 2025; 40(2): 413-427.
  39. Öztürk E, Mesut A, Fidan ÖA. Comprehensive dataset for analyzing detectability in LSB steganography. Bilişim Teknol Derg 2026; 19(1): 79-90.
  40. He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016. pp. 770-778.
  41. Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014.
  42. Huang G, Liu Z, van der Maaten L, Weinberger KQ. Densely connected convolutional networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017. pp. 4700-4708.
  43. Tan M, Le Q. EfficientNet: Rethinking model scaling for convolutional neural networks. In: International Conference on Machine Learning, 2019; PMLR. pp. 6105-6114.
  44. Sandler M, Howard A, Zhu M, Zhmoginov A, Chen LC. MobileNetV2: Inverted residuals and linear bottlenecks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018. pp. 4510-4520.
  45. Dosovitskiy A, Beyer L, Kolesnikov A, Weissenborn D, Zhai X, Unterthiner T, Dehghani M, Minderer M, Heigold G, Gelly S. An image is worth 16×16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020.
  46. Liu Z, Lin Y, Cao Y, Hu H, Wei Y, Zhang Z, Lin S, Guo B. Swin transformer: Hierarchical vision transformer using shifted windows. In: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021. pp. 10012-10022.
  47. Bao H, Dong L, Piao S, Wei F. BEiT: BERT pre-training of image transformers. arXiv preprint arXiv:2106.08254, 2021.
  48. Touvron H, Cord M, Douze M, Massa F, Sablayrolles A, Jégou H. Training data-efficient image transformers & distillation through attention. In: International Conference on Machine Learning, 2021; PMLR. pp. 10347-10357.