Study Of True Random Number Generation Based On Continuous Source

Random number generation, with the features of uniformly distributed, non-periodic, unpredictable and irreproducible, is very important in the applications of in-formation security, statistical sampling and modeling, and deep-space detection. Mostrandom number sequences used in these applications are generated by computer basedon algorithm complexity, which is called pseudo-random number generation (PRNG).PRNGs with deterministic algorithms and short random seeds usually produce uniformlydistributed yet predictable and reproducible random bit sequences with periods in largescale. Therefore, for the applications in information security area, PRNGs may leave po-tential loopholes. Especially, with quantum algorithms and quantum computer designedin hacking area, more and more attentions are increased on the research field of truerandom number generation (TRNG) based on the measurement of physical (quantum)phenomenon.In this dissertation, the author first summarizes the features of PRNG/TRNG and thedevelopments on the research of TRNG. In order to meet the requirements in informa-tion security area, especially for the unconditional security in quantum key distributionsystems, a variety of TRNGs are proposed using discrete random source or continuousrandom source. This dissertation is focused on the theoretical analysis and experimen- tal realization of the continuous random source TRNG based on the quantum feature ofthe laser source/LED source, which are suitable for pursuing super-high random numbergeneration rate in order to meet the requirement of high-speed applications.First of all, the four steps of the implementation of TRNGs are introduced and ana-lyzed in detail, which are physical random source selection, signal sampling and record-ing, data post-processing and random statistical tests. Then, combining with the concreteexperiments, three implementations of TRNGs are presented.1)300Mbit/s TRNG basedon the measurement of the phase noise of a vertical cavity surface emitting laser (VC-SEL);2)40Gbit/s TRNG based on the measurement of the polarization mode-partitionnoise of a VCSEL;3)1.6Tbit/s TRNG based on the measurement of the amplified spon-taneous emission (ASE) noise of a super-luminance LED (SLED). Specifically, the stud-ies include quantum randomness estimation with Shannon entropy and min-entropy, sam-pling rate estimation with the coherent time of the laser source and the overall bandwidthof the experimental setup, analysis and innovation of the post-processing methods ac-cording to data histogram and k-th autocorrelation coefcients, and proposal of additionalstatistical random tests based on the three-standard-deviation model.In the last, the author proposes two indicators, min-entropy and Kolmogorov com-plexity, for the evaluation of the post-processing methods to avoid involving classicalrandomness. With the example of TRNG implementation using VCSEL, the author ap-plies these two evaluation indicators to reveal that inappropriate post-processing wouldmix classical randomness in the final random bit sequence and hence compromise thetrue randomness.