Hybrid Asymmetric Cryptography and Modified Cosine Wavelet Transform-Based Steganography Using Convolutional Autoencoder for Secure Data Sharing

The need for secure data sharing has recently become a critical requirement in modern digital communication networks, especially for image-based information, due to the increasing threats of unauthorized access and privacy breaches. Although cryptography and steganography are widely used for protecting confidential information, it is still a great challenge to achieve high levels of confidentiality and integrity with low computational complexity and minimal error rates. To overcome this challenge, this study proposed the Cosine Wavelet Transformative Convolutional Autoencoded Cryptography Steganography (CWTCA CryptStego) technique, which combines cryptographic and steganographic processes in a Convolutional Autoencoder (CAE) structure. The proposed technique uses Schmidt-Samoa cryptography for secure encryption, Myriad Normalized Residual Filtering (MNRF) for noise removal, and a Modified Discrete Cosine Transform (MDCT) for efficient data embedding, while at the decoder, the stego image is reconstructed with reduced distortion and improved visual quality. Performance analysis using standard evaluation metrics demonstrates that the proposed CWTCA CryptStego technique achieves lower bit error rates, higher Peak signal-to-noise ratio (PSNR), and enhanced confidentiality and integrity, while reducing computation time compared with existing techniques.