Steganography Embedded in Continuous Luminance Modulation for Cover-Source Switching

Low-luminance regions in images often exhibit weak textures, which limits the effectiveness of conventional spatial steganographic methods that rely on local content complexity. More generally, image luminance distributions are strongly influenced by imaging conditions and post-processing operations, among which luminance modulation is widely used in data augmentation pipelines and is natively supported by modern imaging systems. Motivated by the prevalence and physical interpretability of luminance modulation, we propose a novel steganographic framework that integrates message embedding into a continuous luminance modulation process. The framework consists of two key components: (1) a physical-inspired luminance modulation algorithm derived from the atmospheric scattering model, enabling continuously controllable illumination adjustment; and (2) an adaptive embedding scheme that encodes message by controlling the rounding direction of floating-point luminance values, guided by a UNIWARD-like content-aware distortion function. This design embeds information directly during modulation, without requiring additional post-processing steps. By coupling luminance modulation with steganography, the proposed method naturally induces cover-source switching and modulation-parameter variability, which significantly hinders the generalization ability of practical steganalyzers. Extensive experiments demonstrate that the proposed framework provides improved robustness and embedding flexibility under realistic detection assumptions.