All-optical Physical Random Number Generator Based On Laser Chaos

Random numbers have a wide application in scientific calculations such as Monte Carlo simulation, stochastic sampling and artificial neural network. Especially in secure communication, the generation of secure and reliable random numbers, usually called as "KEYs", has a direct bearing on national security, financial stability, trade secrets and personal privacy.The methods of random number generation methods can be classified into two main types:pseudo-random and physical random number generators. Pseudo-random number generators are based on certain algorithms and seeds, can readily generate fast random sequences with rates up to several tens of Gbps and have these merits such as low cost and ease to implement. However, their inherent periodicity will cause serious hidden dangers when they are used in secure communication. Physical random number generators utilize stochastic phenomena in nature as entropy sources to yield unpredictable and aperiodic true random numbers, and thus can guarantee the accuracy in scientific calculation and security in communication. However, conventional physical random number generator are limited at Mbps rates due to the low bandwidth of traditional entropy sources such as thermal noise and oscillator jitter, so it can not satisfy the absolute security of modern high-speed secure communication.Until recently, the appearance of wideband photonic entropy sources such as chaotic laser and amplified spontaneous emission noise greatly promote the rapid development of the physical random number generator. Among them, laser chaos gets great concerns to produce high-speed physical random numbers due to its merits of the high bandwidth and large amplitude. However, all these existing schemes belong to optoelectronic systems, which usually convert fast optical signals into electronic signals, then implement sampling, quantizing and post-processing in the electrical domain to generate high-speed random numbers. To our knowledge, the highest real-time bit rate until now reported physical random generators can reach is4.5Gbps. To enhance this bit rate further, one must select ultra-fast electronic ADCs and XOR gates and thus be bound to face the limitation of "electronic bottleneck", so its improvement room is limited. Although lots of reports point that random number sequences with rates at Gbps or even Tbps can be achieved via wideband photonic entropy sources, but it must be noticed that these ultra-fast bit rates are not real-time on-line rates but only some theoretical expectations obtained through multiplying the trigger clock frequency by the bit number of the multiple-bit ADC.In this thesis, we are devoted to developing novel all-optical physical random number generators, which combine wideband laser chaos sources with all-optical signal processing techniques and implement random bit extraction in the optical domain. This method can efficiently overcome the "electronic bottleneck" encountered by these current optoelectronic systems. Moreover, random numbers generated by the all-optical method can is directly compatible with the current optical communication networks without any electronic-to-optical or optical-to-electronic conversion. Besides, with the development and maturity of the next generation of all-optical networks, this technique is expected to have a greater potential application prospect.Focusing on the research project of "All-optical physical random number generator based on laser chaos", this thesis completes these following works:1. We introduce these application fields and their importance of random numbers and summarize random number generation techniques and their evaluation criteria. Especially, we review in detail those physical random number generators based on chaotic laser are reviewed, show their current study progress according to the international research situation, and finally point out the significance of developing the all-optical random number generation technique.2. We present three different schemes for all-optical random number generation based on continuous-time laser chaos and demonstrate their own feasibility. In the first scheme, we use the wideband chaotic signal from an optical feedback semiconductor laser as the entropy sources, sample the entropy source via the four-wave mixing (FWM) in high nonlinear fibers (HNLFs), employ a HNLF ring oscillator as the all-optical comparator only with the function of comparing, combine the "two-path XOR" technique and finally obtain a5Gbps random bit train. In the second scheme, we use the bandwidth-enhanced chaotic laser from an optically injected semiconductor laser with optical feedback as an entropy source, utilize a Sagnac interferometer based on HNLF as the sampler, employ a λ/4-phase-shifted DFB laser as the all-optical comparator only with the function of comparing, combine the "delay-time XOR" technique and finally obtain a10Gbps random bit sequence. In the third scheme, we also use the bandwidth-enhanced chaotic laser as the entropy source, but employ a Mach-Zehnder (MZ) electro-optical modulator as the sampler and a λ/4-phase-shifted DFB laser as the all-optical flip-flop with two functions of comparing and holding, combine the "delay-bit XOR" technique and finally achieve Gbps random bit generation. All generated random bits can successfully pass these industry-benchmark statistical tests.3. We propose and numerically demonstrate two types of all-optical random number generation methods based on discrete laser chaos. These methods do not need external triggering clocks, sampling and post-processing procedures, but directly quantize the entropy signal into random numbers, and thus reduce the system complexity. Specifically, we in detail analyze and discuss the generation and their stochastic characteristics of these two kinds of discrete-time laser chaos, i.e. the pulse amplitude chaos generated from a passive mode-locked fiber laser and the chaotic self-pulsation from a two-section semiconductor laser with optical injection. Furthermore, we combine them as the physical entropy source with the all-optical flip-flop and numerically achieve the generation of high-speed physical random numbers verified by those industry-benchmark statistical tests.4. We summarize all works in this thesis and point out some possible study direction in the future through combining study hot spots in current high-speed physical random number generators.