The Post-Processing Of Physical True Random Number Generation Based On Chaotic Laser

With the rapid development of electronic commerce, a mass of authentication mechanisms of internet security need massive real-time random numbers. How to produce high speed and plentiful real-time random numbers has paid widely attention. According to the different methods of realization, random number generators can be classified into two categories:pseudo random number generators and physical true random number generators. At present, pseudo random numbers could not satisfy people's requirement for information security. We urgent need high-speed physical true random number generator. Physical true random number generator consists of entropy source, sampler and post-processing. The post-processing is used to eliminate deviation and correlation. Therefore, choosing an appropriate post-processing method directly affects the quality of random numbers. In this dissertation, focusing on the post-processing of physical true random number generation based on chaotic laser, several works are carried out as following: 1. We comprehensively introduce random number generator and the applications. We summarize kinds of present methods of generating pseudo random number and physical true random number.2. The post-processing principle of physical true random number generator is introduced. Chaotic signal generated by optical feedback semiconductor laser is used as entropy signal. We analyze entropy signal by three different post-processing methods, including wavelet denoising post-processing, Von Neumann post-processing and differential post-processing. The result shows that:Using differential post-processing, the secondary peaks of autocorrelation curve are erased. A random number sequence at rates of up to 1 Gbit/s is realized.3. Based on the chaotic source, the results of difference post-processing are analyzed. We investigate the dependence of the randomness test performance on the number of LSBs used and also give the method of determining how many LSBs of the subtractor could be used. We analyzed the reason why NIST SP 800-22 tests can not be passed when LSBs is 3.