Dynamic Chaotic–Adversarial Framework for High-Capacity and Imperceptible Image Steganography

Abstract

The rapid growth of digital communication has intensified concerns regarding data confidentiality as sensitive information transmitted through multimedia images is increasingly vulnerable to interception and unauthorized analysis. Conventional image steganography methods often struggle to simultaneously achieve high embedding capacity, strong imperceptibility, and resistance to modern steganalysis. To address this challenge, this study proposes a steganographic framework that integrates dynamic logistic chaotic encryption with an adversarial feature-level embedding network. The chaotic sequence is generated using a time-varying logistic map within a highly unstable region, where the control parameter is adaptively derived from a hash-modulated process to produce unpredictable keystreams and strengthen payload security. The encrypted secret image is then embedded through a GAN-based generator guided by a discriminator to preserve natural image characteristics, while a dedicated extractor ensures accurate recovery. Experimental results on multiple standard test images with resolutions of 256 × 256 and 512 × 512 demonstrate high visual fidelity, achieving PSNR values above 58 dB and SSIM values above 0.995, supported by nearly identical histogram distributions between cover and stego images. These findings indicate that the proposed framework provides a promising solution for secure multimedia communication by enabling visually imperceptible and reliably recoverable hidden transmission in digital images.