How to know if your random number generator is really random

Terry Gerton So let’s start with the basics. Why is it so important to generate truly random numbers?

Travis Humble Random numbers are a fascinating topic. Of course, they represent what would be an unexpected sequence of digits: things of one, two, and three, but in random order. It turns out that we use randomness around us in many different places. Some of those is in encryption, which we use for securing information. If you have a bank account or an email, you most likely are using some form of what we call public key encryption. Which transmits information back and forth, but in a protected state. That encryption methodology actually makes use of random numbers in order to secure that information. And it’s the randomness in the number that makes it protected and secure. If someone was able to anticipate or perhaps guess what the number was, then it would compromise the security of those systems. And in this case, by having numbers that are truly random, you can mitigate any concern that someone would be able to guess them. And ultimately get into what might be considered a secret or sensitive information.

Terry Gerton Well, that’s a really helpful framework to then talk about, more specifically, how you generate random numbers, because you just implied that some trusted people can generate random numbers, and maybe some untrusted people generate random numbers.

Travis Humble Generating random numbers is actually an art in and of itself. The task of creating an absolutely random sequence is actually really difficult when you stop and think about it. None of us individually would be especially good at it. We all have some sort of biases either on experience or environment that can lead us to making choices that a really savvy adversary might be able to anticipate. But the natural world actually has similar challenges as well. If you are trying to use random fluctuations of light, or heat, or other sorts of physical quantities, those can often be translated back to some cause and effect relationship, and if you know the cause, you can deduce what the effect’s going to be. But there is one place where that’s not true, and this happens in quantum mechanics. Quantum mechanics, at the best of our ability, is a process in nature that is fundamentally random. Effectively, we don’t know what’s going to happen sometimes when we’re looking at the quantum mechanical systems. And if we can harness that, we can actually create truly random sources of generating number sequences that can then support applications like the cryptography that we were talking about.

Terry Gerton So in the current process for generating random numbers, there might be an opportunity to guess the next random number.

Travis Humble You’re exactly right. Right now, because many of the technologies that we use are based on what we refer to as classical physics or classical mechanics, we actually have to mimic the process of randomness. There are very clever algorithms, which we refer to as pseudo-random number generation algorithms, that are responsible for creating sequences of numbers in a way that makes it appear as if they are random. There’s tests that you can apply to these random sequences, and to really high standards, you can assume that they are random for most purposes. But again, the risk of them being deduced or discovered can oftentimes raise the concern and raise the threshold that you want those pseudo-random numbers to pass. In this case, the quantum mechanical processes we’re talking about don’t suffer from that defect. They actually will always truly be random and be able to pass those tests regardless of the tolerances.

Terry Gerton So I think I’m getting closer to it. Our current processes are random, but not random enough, and you and your team have experimented with a new protocol for generating and certifying randomness. So what’s different about your approach?

Travis Humble Fundamentally, it comes back to this quantum-mechanical process. Whether we’re talking about the individual particles that make up molecules or matter or other forms of materials, those are quantum- mechanical. And when we start to probe those systems, we encounter this randomness phenomenon that we’re taking about. Most recently, we have begun to harness that into a new type of technology which we refer to as quantum computers. These are information processing systems that actually use those same quantum mechanical processes in order to store and calculate results. One of the most straightforward calculations that you could perform is a calculation to generate randomness. Basically, think of this as flipping a coin and trying to create a string of heads or tails that come out from multiple times of flipping that coin. In this particular case, though, we’re not using a coin. We’re actually using individual electrons and other types of subatomic particles whose state cannot be anticipated prior to the calculation. In this case, we actually use the quantum computer to generate that randomness and then can test it to show that it passed all the requirements that we have for being a truly random number.

Terry Gerton I’m speaking with Dr. Travis Humble. He’s the director of the Quantum Science Center at Oak Ridge National Laboratory. All right, so now you’ve got truly random, certifiable random numbers. What do you do with those?

Travis Humble Lots of different things. One of the first things, though, which we focused on is checking to see if they really are truly random. The tests that we were talking about earlier, those of course apply to the random sequences themselves, but the process of generating them can be cast into question, especially when we’re using a new technology like quantum computing. In this case, we actually use the high-performance computing systems at Oak Ridge National Laboratory as a way of confirming that there was no computationally feasible method of creating the same strings of random numbers that the quantum computer created. That was, of course, a very significant test. It did require close collaboration between our facilities here at Oak Ridge, as well as our industry partners at JPMorgan Chase and Quantinuum, who were creating the random numbers on their side. In this case, the first thing we did was actually compare the results of trying to do those calculations both quantum mechanically and classically. And we found, of course, that the quantum mechanical version was not only a very fast way of creating random numbers, but they were also a genuinely and truly random sequence. Now that these types of random numbers have been created, you can imagine using that as a service for many different types of applications. One of the ones that we’re most interested at Oak Ridge is actually using randomness as input to the types of scientific calculations we do when we’re looking at novel materials or high-energy reactions in fusion or fission reactors. What happens there is that many times we need to randomly sample what we think would be the outcome from those events and average the statistics in order to get good estimations and predictions of how those systems will behave. Having truly random numbers gives us an unbiased estimate of these types of statistics that can be incredibly powerful. Now, of course, there’s other applications as well, and our friends in industry are especially interested in the cryptography application, the application for using those simulation techniques in financial modeling, as well as many other areas.

Terry Gerton So you’ve completed the experiment. Is the technology, the process now deployable or does it still need more experimentation?

Travis Humble I think the experimentation itself as a science project has actually gone a very long way and I have a lot of confidence in the results from this particular experiment. Translating these quantum computers into services though continues to be an industry challenge. Basically, what I’m trying to say is that we have these quantum computers today as prototype systems that can be used for testing and evaluating some of these novel protocols. I do anticipate in the future that we will be able to deploy these types of quantum computers including at national laboratories and financial institutions and many other locations that could be used for exactly these types services. Think of it as an online supplier of random numbers that you might want to leverage for any number of applications.

Terry Gerton And so you mentioned that industry is very interested in moving this forward into practice. What do you think the timeline is?

Travis Humble I actually think that we can be very aggressive in this area. There is a lot of incentive at the moment to keep data secure, and of course, coming up with novel encryption methods that can be truly, provably secure would be a powerful capability in this space. In that regard, I think one of the main challenges right now might be raising awareness about what are the opportunities available, both from the conceptual side, the science that we’ve been developing here, but then also from the applications and product side. I know many of our industry partners are still trying to define their application space, create the market for these new technologies. And I think that is actually moving ahead very quickly because of the high value and potential that it has.