Context-aware tactical image encryption: toward exact (0.0 MSE) reconstruction via integer wavelets and biological DNA algebra
Year: 2026
Type: article
Abstract: This paper presents a context-aware image-encryption framework for high-resolution tactical imagery, built around two central, empirically demonstrated properties. First, an Integer Haar Lifting Scheme enables exact, bit-for-bit lossless integer-domain reconstruction, avoiding the floating-point rounding error inherent to standard wavelet transforms. Second, a Master-Slave DNA-encryption stage combined with an invertible cross-band wave-diffusion round achieves strong ciphertext diffusion under both plaintext-bound and fixed-key conditions; key generation is driven by a three-dimensional Lorenz chaotic system with plaintext-bound, channel-specific PBKDF2 derivation. The framework additionally incorporates a context-aware Master-Slave routing mechanism that reads the Low-Low (LL) wavelet sub-band to apply heavy DNA-Addition diffusion only to structurally important regions, downgrading to lightweight DNA-XOR elsewhere. This routing design is architecturally motivated by the goal of reducing computation on power-constrained edge devices such as UAVs, but that efficiency benefit is currently theoretical rather than empirically confirmed: a dedicated routing ablation (Section 4.8 ) found no measured wall-clock advantage of adaptive over uniform DNA routing across three independent test machines in the current NumPy-vectorized implementation. On a representative tactical Synthetic Aperture Radar (SAR) image, the proposed framework achieves 7.9998 bits of Shannon entropy and exact bit-for-bit lossless reconstruction (MSE = 0.0, PSNR = ∞, SSIM = 1.0) under the compatible numerical environment used in these experiments, with mean NPCR of 99.6091% and UACI of 33.46% across 10 test images under plaintext-bound key derivation. A fixed-key differential test initially exposed near-zero intrinsic ciphertext diffusion (NPCR ≈ 0.0003%); we address this directly with the invertible cross-band wave-diffusion round mentioned above, raising fixed-key NPCR to 99.71% (99.79% ± 0.13% over 20 trials) and UACI to 34.01% (33.61% ± 0.59%) while preserving exact losslessness and near-zero ciphertext correlation. We report these results alongside their limitations: in its current unoptimized implementation, the proposed framework runs approximately 12–28 × slower than a pure-Python AES-256-CBC baseline, and, as noted above, the adaptive-routing efficiency premise is not yet empirically supported. Broader dataset evaluation, formal active-attack assessment, and cross-platform reproducibility verification remain as prioritized future work.
Institution: Vellore Institute of Technology University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,990
Funder: VIT University
OpenAlex ID: W7236845951
