Secure imaging in dynamic scattering channels via hyperchaotic-encrypted multimodal diffusion with adaptive quantization

In dynamic scattering environments, the secure generation and transmission of images are hindered by issues such as signal distortion and data loss, which degrade encrypted image quality and transmission efficiency. This study proposes a multimodal diffusion network integrating hyperchaotic systems and quantization to address secure image transmission challenges in dynamic scattering environments. The proposed method utilizes hyperchaotic mapping for nonlinear encryption and quantization for efficient compression, resulting in a 0.7 improvement in compression ratio and a 50% enhancement in transmission rate while maintaining image fidelity (PSNR = 20.1 dB, SSIM = 0.52 at 5 dB SNR). The adversarially trained network demonstrates robust adaptability against multimodal attacks, showing 18.8%, 7.7%, 19.7%, 26.4%, and 29.3% anti-interference improvements over single-mode methods for Gaussian noise, pulse interference, band loss, simulated/real dynamic scattering attacks, respectively. The framework ensures global stability in time-varying channels through adaptive strategy optimization, effectively balancing structural integrity with encryption efficiency. Experimental findings substantiate its aptitude to mitigate signal distortion and data loss while preserving security-critical features under complex scattering conditions.