HyCrySteg-BFO: Crypto-Steganography Hybrid Model Leveraging modified UNet Architecture and Bacterial Foraging Optimization (BFO)

Abstract

In the face of increasing data breaches over various organizations, traditional security methods are proving insufficient to protect sensitive information. This paper introduces a novel Crypto-Steganography system that combines the strengths of cryptography and steganography to enhance data security. The proposed model leverages a modified UNet architecture for precise background segmentation, Integer Wavelet Transform (IWT) for efficient data embedding, and Bacterial Foraging Optimization (BFO) for optimal pixel selection. Additionally, fractal images are employed as cover images to enhance robustness. Moroever, the encryption process is divided into two phases-confusion and diffusion—utilizing chaotic maps to ensure high security with minimal latency. The modified UNet architecture, enhanced with GhostBottleneck Residual blocks (GB_RB) and switchable normalization, demonstrates superior performance in background segmentation, achieving high accuracy across diverse datasets. The proposed modified Unet model achieved an improvement in Dice coefficient up to 8% in BUID medical dataset and up to 6% in Kvasir dataset. Additonally, Extensive evaluations on various benchmark datasets demonstrate that the proposed model outperforms state-of-the-art methods in terms of embedding capacity, imperceptibility, and robustness against attacks. key findings from extensive evaluations on various benchmark datasets reveal that the proposed model significantly outperforms existing methods in terms of imperceptibility reaching improvement up to 5% db in PSNR.