A SHA-256 driven multi-PWLCM image encryption scheme with adaptive image-dependent permutation-diffusion

Abstract With the rapid growth of digital image transmission over open and insecure communication channels, the need for robust and computationally efficient image encryption techniques has become increasingly important. This paper presents a SHA-256-driven multi-PWLCM image encryption scheme with adaptive image-dependent permutation–diffusion for secure visual data protection. Rather than constructing a new chaotic system, the proposed framework adopts three conventional Piecewise Linear Chaotic Map (PWLCM) instances within a plaintext-dependent, functionally role-separated architecture. Plaintext-derived SHA-256 information diversifies the operational initial conditions and control parameters of the three PWLCMs, while the PWLCM instances are explicitly assigned to row permutation, XOR diffusion, and column permutation, respectively. An adaptive hash-dependent flipping mechanism further introduces image-dependent spatial transformation, while the same processing framework supports both grayscale and color images. The security behavior is evaluated through histogram and chi-square analysis, adjacent-pixel correlation, global and local information entropy, differential analysis, key sensitivity, chosen-plaintext and ciphertext-based cryptanalytic evaluations, noise and cropping robustness, and statistical randomness testing. Direct randomness evaluation of the PWLCM-generated sequences and finite-precision analysis under IEEE-754 double-precision arithmetic are additionally performed to assess the numerical behavior of the underlying chaotic generators. The experimental results show near-uniform ciphertext distributions, information entropy close to the theoretical value for 8-bit data, low adjacent-pixel correlations, and NPCR and UACI values close to their expected levels. Chosen-plaintext experiments further show substantial plaintext-dependent changes in the generated row-permutation indices, column-permutation indices, and diffusion-key elements, limiting direct reuse of the tested equivalent permutation–diffusion information across different plaintexts. Computational analysis indicates that the principal pixel-wise operations are linear, while sequence sorting determines the conservative overall complexity bound of $$O(P\log P)$$ . Overall, the results demonstrate the feasibility of combining plaintext-dependent parameter diversification, explicit functional separation of lightweight PWLCMs, and unified grayscale/color processing within a comparatively compact image-encryption architecture.