Automation of reversible steganographic coding with nonlinear discrete optimisation

Authentication mechanisms are at the forefront of defending world from various types cybercrime. Steganography can serve as an authentication solution through use a digital signature embedded in carrier object to ensure integrity and simultaneously lighten burden metadata management. Nevertheless, despite being generally imperceptible human sensory systems, any degree steganographic distortion might be inadmissible fidelity-sensitive situations such forensic science, legal proceedings, medical diagnosis military reconnaissance. This has led development reversible steganography. A fundamental element steganography is predictive analytics, for which powerful neural network models have been effectively deployed. Another core coding. Contemporary coding based primarily on heuristics, offers shortcut towards sufficient, but not necessarily optimal, capacity--distortion performance. While attempts made realise automatic with networks, perfect reversibility unattainable via learning machinery. Instead relying heuristics machine learning, we aim derive optimal by means mathematical optimisation. In this study, formulate nonlinear discrete optimisation problem logarithmic capacity constraint quadratic objective. Linearisation techniques developed enable iterative mixed-integer linear programming. Experimental results validate near-optimality proposed algorithm when benchmarked against brute-force method.