Abstract
<title>Abstract</title> <p>The exponential growth of heterogeneous digital assets demands robust, format-independent encryption frameworks capable of securing universal binary data. Existing matrix-based cryptosystems predominantly target specific modalities like images, utilize small, fixed matrix dimensions, and suffer from complex key storage or padding-induced security vulnerabilities. To address these limitations, this paper proposes a scalable hybrid encryption framework that processes raw byte streams directly, enabling universal protection for diverse digital file formats. The proposed architecture integrates three complementary mechanisms: a user password-derived Logistic Map for pseudo-random chaotic sequence generation, a scalable Unimodular Hill Cipher that constructs large invertible key matrices dynamically without repeated determinant testing, and the Advanced Encryption Standard in Cipher Block Chaining (AES-CBC) mode to securely encrypt residual data segments. Furthermore, an entropy-driven selection strategy is introduced to adaptively adjust the matrix dimensions based on file size and input statistical characteristics, balancing cryptographic strength and computational efficiency. Experimental evaluations across various binary datasets demonstrate near-ideal Shannon entropy approaching 7.9998 bits/byte, negligible adjacent byte correlation, and an avalanche effect close to the 50% theoretical optimum. The framework passes the NIST SP800-22 randomness tests and guarantees 100% lossless data recovery via MD5 integrity verification. Consequently, the proposed cryptosystem successfully eliminates structural redundancies and provides strong resistance against statistical, differential, and entropy-driven cryptanalytic attacks, making it a highly practical and scalable solution for universal binary file security.</p>