Оцінка ентропії та фрактальних властивостей псевдовипадкових генераторів чисел для шифрування

Abstract

This work is devoted to the evaluation of pseudorandom number generators in the field of cryptography, in particular their entropy and fractal properties. The study proposed a hybrid generation algorithm that combines a linear congruent method with material sources of microcontroller randomness - ADC thermal noise and uninitialized SRAM memory states. Shannon entropy and minimal entropy were used to estimate the unpredictability coefficient. Fractal analysis, in particular the calculation of the Hearst indicator, is given special attention to identify hidden patterns in the data structure. The hybrid algorithm was compared with its base models using the NIST SP 800-22 test suite for statistical validation. The results obtained suggest that the combination of physical noise and the linear congruent method provides a balanced approach to safety and computational speed. This approach is effective for microcontrollers, which are weak and provide a reliable basis for secure data transmission in IoT networks.