A Comparative Study on the Robustness of Chaos-Based Random Number Generators

This paper presents a comparative study on continous-time chaos-based random number generation methods regarding their robustness against changes in chaos controlling parameters and external interference. Chaotic systems suggest enabling high throughput random data without need for postprocessing and with less complex hardware. However, due to effects of aging or fabrication process variations, the chaos controlling parameters of the random number generator may change. Furthermore, external interference can be applied on the chaotic oscillator to manipulate its output. Therefore, in a chaotic RNG, the bit generation method should be immune to parameter variation and external interference. In this study, two widely used methods for random number generation have been compared: 1) Regular sampling of chaotic waveform (RSCW), and 2) Chaotic sampling of regular waveform (CSRW). A double-scroll chaotic system is chosen as the chaotic oscillator and it is numerically simulated in normalized time domain to generate random bit sequences using both methods. Applying the concepts of autocorrelation and approximate entropy to the output bitstreams, the robustness of the two bit generation methods against parameter variation and external interference have been compared. It is demonstrated that chaotic sampling of regular waveform method provides more robustness against parameter changes and external interference compared to regular sampling of chaotic waveform method.