STEGOCRYPTNET: A DEEP LEARNING AND NTRU-BASED CRYPTOGRAPHIC IMAGE STEGANOGRAPHY FRAMEWORK

Abstract

In present digital age, protecting and safeguarding essential information against unauthorized access is considered crucial. Thus, image steganography is regarded as they offer strong approach fo concealing the secret data within images, allowing secure communication. Henceforth, this paper focuses on developing a SecureNet-Stego method for achieving improved image steganography process by integrating Deep Learning (DL) model with advanced cryptographic techniques. The proposed model is developed using Discrete Wavelet Transform (DWT) for attaining optimized frequency domain embedding while enhancing both imperceptibility and robustness. To further attain increased security, 𝑁𝑡ℎ degree polynomial Ring Unit (NTRU) based encryption algorithm, a post-quantum cryptographic technique which encrypts the secret image prior to embedding to protect it against cryptanalysis attacks. Moreover, SecureNet-Stego model is deployed using a morphological Transformer based Generative Adversarial Network (GAN) for achieving precise and accurate encoding and extraction of secret image. The proposed system is implemented using python software and the obtained results showcases improved steganography capacity, resilience and enhanced security and reliability, leading to highly efficient performance with secure digital communication.