With the increased number of personal electronic devices, the security of these devices and personal information has become a big concern. As effective solutions to securing a system, encryption and the signing of confidential information with digital key streams have been widely used. These digital key streams are usually generated by random number generators (RNGs), which can be classified into true random number generators (TRNGs) and pseudo-random number generators (PRNGs). The difference is that TRNGs derive randomness from analog physical processes, such as thermal noise, voltage noise, and so on; PRNGs derive randomness from computational complexity, which is decided by the seed. This paper focuses on TRNGs, which are usually used in systems requiring a high level of security. The mechanism behind the on-chip TRNGs is that the circuit converts all of the transistor-level randomness (like random telegraph noise, flicker noise, and thermal noise) into voltage or delay signals. A ring oscillator is one of the best-fitting circuits for this purpose; it is simple to build and also easy to control. Most importantly, it can capture noise accurately. Although designing a ring oscillator is easy, the challenge of such a TRNG is how to evaluate it. To make it clearer, since TRNGs deal with randomness, which is really hard to predict and verify, it is very hard to evaluate and test TRNGs.

In this paper, Lao et al. provide a comprehensive study of the randomness of different variation sources, and propose a model for beat frequency detector-based high-speed TRNGs (BFD-TRNGs). The model is calibrated against the actual measurement. According to the paper, “the key contribution of the proposed approach lies in fitting the model to measured data and the ability to use the model to predict performance of BFD-TRNGs that have not been fabricated.” Based on the proposed model, the authors also present several novel designs of BFD-TRNGs, such as parallel BFD based, cascade BFD based, and a combination of both. The paper compares the pros and cons of each and also validates the model.

This paper provides a very useful TRNG model for researchers working in the hardware security area. Also, the comparison study of different TRNGs will guide researchers to pick the right one to satisfy different requirements and metric budgets.