VisionStego: An AI-Guided Dual-Layer Cryptographic Steganography Framework for Secure Cloud Data Transmission

Abstract

Modern cloud platforms now carry a substantial share of the world\'s sensitive digital traffic, and defending that traffic calls for more than hiding its meaning - an intercepted ciphertext is still visibly a target worth attacking. Encryption alone protects content; it does nothing to disguise the fact that a secret is being sent at all. This paper proposes VisionStego, a dual-layer security architecture that pairs symmetric-key encryption with an artificial-intelligence-guided steganographic embedding stage, so that cloud-hosted data is protected in both substance and appearance. The secret payload is first reduced in size through wavelet-based compression, then encrypted with a shared symmetric key, and finally concealed inside a cover image at positions chosen by a trained Convolutional Neural Network (CNN) rather than by a fixed or pseudo-random rule. The network scores each candidate pixel for embedding suitability using edge strength, local variance, and texture complexity, concentrating modification in regions where it is least visible to the eye and least anomalous to statistical steganalysis. Tested on the Lena, Baboon, and Peppers benchmark images, the resulting stego images reach a Peak Signal-to-Noise Ratio (PSNR) of up to 46.2 dB and a Structural Similarity Index (SSIM) of 0.985, exceeding conventional LSB, adaptive LSB, and wavelet-based baselines on every image and every metric tested. These findings indicate that VisionStego offers a practical route to the combined concealment and confidentiality that neither cryptography nor steganography can deliver in isolation.

Introduction

This paper presents VisionStego, a cloud security framework that combines cryptography, steganography, and artificial intelligence to protect sensitive data in cloud environments. While traditional encryption methods such as AES, RSA, and ECC secure data by making it unreadable, they do not hide the existence of encrypted information. Steganography addresses this limitation by concealing encrypted data within digital images. However, conventional Least Significant Bit (LSB) steganography is vulnerable to statistical and machine-learning-based detection because it embeds data without considering image characteristics.

VisionStego overcomes these limitations by using a Convolutional Neural Network (CNN) to identify image regions with high texture and edge complexity, where hidden data is less likely to be detected. The framework first compresses the secret payload using the Discrete Wavelet Transform (DWT), encrypts it with a symmetric key, and then embeds the encrypted data into the selected image pixels using adaptive LSB substitution. This produces a stego image that appears visually identical to the original while keeping the hidden data both concealed and encrypted.

The paper identifies several research gaps in existing hybrid cryptography–steganography systems, including fixed or pseudo-random pixel selection, poor integration between compression, encryption, and embedding stages, vulnerability to modern steganalysis techniques, limited evaluation across different images, and insufficient consideration of real-time cloud performance requirements. To address these issues, the proposed framework integrates all security components into a single coordinated pipeline.

The objectives of the research are to develop a two-layer cloud security architecture, design a CNN-based adaptive pixel selection model, incorporate wavelet compression, evaluate performance using standard image-quality metrics such as PSNR, MSE, and SSIM, and compare the proposed method with conventional steganographic techniques.

The literature review shows that previous studies have successfully combined encryption and steganography for cloud security but often suffer from high computational overhead, limited resistance to machine-learning-based steganalysis, or poor embedding efficiency. Recent deep learning approaches improve image quality and robustness but are frequently computationally expensive or focus only on detection rather than secure embedding.

VisionStego addresses these shortcomings by integrating wavelet compression, symmetric encryption, CNN-guided adaptive pixel selection, and adaptive LSB embedding into a unified architecture. The system consists of a Sender Module, which performs compression, encryption, and embedding, and a Receiver Module, which extracts, decrypts, and reconstructs the hidden data using the same CNN model and shared secret key. Overall, the framework aims to provide a secure, efficient, and practical solution for protecting cloud-based communications while maintaining high image quality and resistance to modern steganalysis techniques.

Conclusion

This paper has presented VisionStego, an AI-based two-layer security architecture that combines cryptographic protection with intelligent steganographic embedding for cloud data security. It addresses the principal shortcomings of existing hybrid frameworks by tying together four coordinated components: Discrete Wavelet Transform compression to shrink the payload, symmetric-key encryption for confidentiality, CNN-based adaptive pixel selection for imperceptibility, and LSB substitution for the embedding itself.On standard benchmark images, VisionStego reaches PSNR values of up to 46.2 dB and SSIM values of up to 0.985, well ahead of Traditional LSB, Adaptive LSB, Wavelet-based, and CNN-based baselines. The qualitative comparison confirms Very High security and image quality alongside Very Low detectability, supporting its suitability for real-world cloud communication scenarios.More broadly, the results suggest that intelligently combining machine-learning-based pixel selection with established compression and encryption techniques produces a security improvement that no single component could deliver alone - a synergistic effect rather than an additive one. VisionStego is offered as a practically deployable contribution toward that combined goal.

References

[1] Angulo, K., Gil, D., Yáñez, A., Espitia, H.: Hybrid CNN–NLP model for detecting LSB steganography in digital images. Applied System Innovation 8(2), 1–18 (2025). [2] Badhan, A., Malhi, S.S.: Enhancing data security with hybrid cryptography and steganography. In: Proc. 2nd Int. Conf. on Advances in Computation, Communication and Information Technology (ICAICCIT), Faridabad, India, pp. 1247–1252 (2024). [3] Cheng, Y., Zhou, J., Chen, J., Yin, Z., Zhang, X.: Robust fixed neural network steganography with popular deep generative models. IEEE Trans. Information Forensics and Security 20, 1102–1115 (2025). [4] Farooq, N., Mir, R.: Image steganalysis using deep convolution neural networks: a literature survey. Int. J. Sensors, Wireless Communications and Control 14(1), 1–22 (2024). [5] Goyal, A., Gupta, B.: A hybrid approach using cryptography and K-least significant bit steganography algorithm in cloud computing. In: Proc. 1st Int. Conf. on Advances in Computer Science, Electrical, Electronics, and Communication Technologies (CE2CT), Bhimtal, India, pp. 196–202 (2025). [6] Gupta, D., Verma, A.: Machine learning-based steganography detection and prevention techniques. Computers & Security 138, 103–119 (2024). [7] Hassan, S., Muztaba, M.A., Hossain, M.S., Narman, H.S.: A hybrid encryption technique based on DNA cryptography and steganography. In: Proc. IEEE 13th Annual Information Technology, Electronics and Mobile Communication Conf. (IEMCON), Vancouver, Canada, pp. 501–508 (2022). [8] Huang, K., Liu, J.: High-capacity image steganography using deep residual networks. IEEE Trans. Circuits and Systems for Video Technology 34(5), 3412–3424 (2024). [9] Jain, P., Saini, A., Singh, A., Shreya, S.: Secure data transmission with steganography. In: Proc. 11th Int. Conf. on Reliability, Infocom Technologies and Optimization (ICRITO), Noida, India, pp. 1–4 (2024). [10] Kanimozhi, R., Padmavathi, V.: Robust and secure image steganography with recurrent neural network and fuzzy logic integration. Scientific Reports 15, article 4821 (2025). [11] Khalifa, A., Yadav, Y.: Wavelet-based fusion for image steganography using deep convolutional neural networks. Electronics 14(14), 1–18 (2025). [12] Kumar, M., Singh, R.: Secure image steganography using deep learning and edge detection. J. Information Security and Applications 80, 103–117 (2024). [13] Lee, H., Kim, J.: Deep learning-based steganography for secure image communication. IEEE Trans. Multimedia 26, 5821–5834 (2024). [14] Li, S., Zhang, X., Zhang, W.: Image steganography and style transformation based on generative adversarial network. Mathematics 12(4), 1–17 (2024). [15] Liu, F., Zhang, T., Zhang, C.: CLPSTNet: a progressive multi-scale convolutional steganography model integrating curriculum learning. Pattern Recognition Letters 182, 44–52 (2025). [16] Malathi, P., Kumar, T.G.: Deep steganographic approach for reliable data hiding using convolutional neural networks and adaptive loss optimization. Scientific Reports 15, article 6103 (2025). [17] Patel, A., Mehta, S.: AI-driven adaptive LSB steganography for secure multimedia communication. Int. J. Computer Applications 185(22), 1–8 (2023). [18] Prakash, A., Chauhan, S.: Exploring innovative methods for enhancing data security in computing. In: Proc. 3rd Int. Conf. on Smart Data Intelligence (ICSMDI), Trichy, India, pp. 63–68 (2023). [19] Rahman, S., Uddin, J., Hussain, H.: A novel and efficient digital image steganography technique using least significant bit substitution. Scientific Reports 15, article 2341 (2025). [20] Rao, A.S., et al.: A secured cloud architecture for storing image data using steganography. In: Proc. 2nd Int. Conf. on Computer, Communication and Control (IC4), Indore, India, pp. 1–6 (2024). [21] Reddy, K.N., Snehitha, M.S., Reddy, K.S.C., Reddy, K.D., Rajagopal, S.M., Panda, N.: Secured steganography: leveraging cryptography in a cloud-based toolbox. In: Proc. 7th Int. Conf. on Contemporary Computing and Informatics (IC3I), Greater Noida, India, pp. 144–150 (2024). [22] Roy, S., Das, K.: Hybrid steganography model combining wavelet transform and deep learning. Multimedia Tools and Applications 83(15), 45123–45141 (2024). [23] Saxena, A.K., Sinha, S.K.: Design and development of image security technique by using cryptography and steganography: a combine approach. Int. J. Image, Graphics and Signal Processing 10(4), 13–21 (2018). [24] Sharma, A., Gupta, P.: Adaptive image steganography using machine learning-based pixel selection. IEEE Access 12, 55321–55334 (2024). [25] Sharma, R., Patel, V.: Secure image communication using AI-assisted steganography. Multimedia Systems 29(3), 1421–1435 (2023). [26] Shidaganti, G., Vinay, M., Patil, P.: Enhancing data protection using cryptography and image steganography in cloud environment. In: Proc. 5th Int. Conf. on Circuits, Control, Communication and Computing (I4C), Bangalore, India, pp. 93–99 (2024). [27] Singh, P., Gupta, R.: Deep learning approaches for steganography and steganalysis: a review. Artificial Intelligence Review 57(4), 1–38 (2024). [28] Sujithra, T., Justus, V., Mahaboob, M.I., Durai, S.: Advanced security model for cloud infrastructure using cryptography and steganography. In: Proc. Int. Conf. on Intelligent & Innovative Practices in Engineering & Management (IIPEM), Singapore, pp. 1–7 (2024). [29] Ullah, A., Haque, M.I., Hossain, M.M., Ahammad, M.S., Aktar, M.N.: A novel LSB steganography technique for enhancing cloud security. J. Information Security 15(3), 355–377 (2024). [30] V., A.K., Nathiya, R.: Cryptography and steganography fusion for robust cloud file protection. In: Proc. Int. Conf. on Inventive Computation Technologies (ICICT), Kirtipur, Nepal, pp. 1442–1446 (2025). [31] Wang, L., Liu, Y.: Adaptive image steganography using generative adversarial networks. IEEE Signal Processing Letters 30, 1401–1405 (2023). [32] Yang, H., Xu, Y., Liu, X., Ma, X.: PRIS: practical robust invertible network for image steganography. In: Proc. IEEE Int. Conf. on Computer Vision (ICCV), Paris, France, pp. 10745–10754 (2023). [33] Yu, J., Zhang, X., Xu, Y., Zhang, J.: CRoSS: diffusion model makes controllable, robust and secure image steganography. Advances in Neural Information Processing Systems 36, 1–13 (2023). [34] Zhang, S., Xiao, Y., Tian, H.: A multi-image steganography method for secure information hiding. Cybersecurity 8(1), 1–15 (2025). [35] Zhang, T., Chen, X.: Secure image steganography using CNN-based feature extraction. J. Visual Communication and Image Representation 92, 103–116 (2023). [36] Zhao, M., Wang, J., Xu, H.: Deep neural network-based image steganography with improved payload capacity. IEEE Trans. Information Forensics and Security 18, 3101–3112 (2023).

Copyright

Copyright © 2026 Rajiv Saxena, Priti Maheshwary. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.