Abstract
Random number generator (RNG) is an important part of communication and protocols in cryptography. It generates the cryptography keys and initial numbers, and also could be used in any application that needs the random value. It is hard to define a good RNG design by the unpredictable character or by the statistical test result. To be unpredictable, the analog RNG is chosen. However, the influence from different environments would cause uncontrollable quality of RNG. To be good in statistical test result, the digital RNG is chosen. Thus the random outputs are generated from the iterative computations of the chosen equation with good statistical feature, and the outputs are working on the deterministic curve. A good statistical result depends on the chosen equation and on the highly precise digital operation. In this thesis, a chaos-based RNG for a shorter bit length of 32 in pure digital environment with pseudo noise influenced is proposed. Once the pre-decided timing (ranged from 600 to 800 in this work) arrives, some of the least significant bits would be set to fixed values called the pseudo noises and feedback to the chaos equation operated. In the algorithm implemented, the proposed RNG would have the analog and digital feature simultaneously. By tuning the timing of reseeding, an acceptable quality of randomness is obtained without operating in hi-precise digital operation, exclusive-or gates or the parity filter to scramble the output.