Researchers from KAUST (King Abdullah University of Science and Technology) have recently unveiled a revolutionary manufacturing method for "memristors" - integral circuits among the fundamental electrical components including resistors, capacitors, and inductors. This cutting-edge technique has demonstrated its potential to facilitate the development of a crucial cryptographic element, namely a True Random Number Generator (TRNG).
Although it may seem counterintuitive (after all, how difficult is it to generate random numbers? ), true random number generators are crucial components of cryptography and also one of the most prone to failure. This is due to the ease with which a random distribution—where each possible event has an equal chance of occurring—can transform into a non-random distribution.
TRNGs are typically implemented on silicon, like AMD's Ryzen and Epyc-bound Cryptographic Co-Processor (CCP), which is currently at iteration 5.0. Examining naturally random occurrences, such as the photoelectric effect that underlies the operation of our computers, is one technique to get random numbers. These results lead to the generation of random numbers, which are then used as the foundation for an encryption procedure. This process is known as hashing, and it involves converting each random number into a portion of the encrypted information. Consider the fact that AMD's Xilinx business sells Field-Programmable Gate Arrays (FPGAs) with the intention of acting as True Random Number Generators to properly put the issue into context.
However, electrical components have operational limits, and slight voltage fluctuations can generate "errors" that take the shape of patterns in calculation or photoelectricity. Naturally, when patterns appear among a set of numbers that are intended to be random, the numbers are no longer truly random. There is a pattern, and the likelihood that one number will be chosen over another varies significantly. If it isn't really random, the emergent patterns can be extracted, examined, and measured against the output that was encrypted. The path to the purportedly secure message using cryptography is also exposed.
Certain patterns can arise organically when imbalances occur within a system, causing it to deviate from its random "equilibrium" state. This phenomenon, observed in hardware degradation over time, contributes to the decline in maximum sustained operational frequency in both CPUs and GPUs as they age. Researchers have capitalized on these patterns, extracting information by analyzing indicators like system fan speed. However, more advanced adversaries can intentionally introduce patterns of their own.
The KAUST researchers' innovation has now made it possible to fabricate memristor-based TRNG in a manner akin to 3D printing. Instead of the typical filament, boron nitride and silver electrodes are formed in atomically thin layers, building up the components of a memristor one on top of the other. The TRNG consumes less power as a result of this unique production process than the typical CPU-integrated alternatives, which are constructed from expensive circuits with millions of transistors (expensive both in terms of the power they consume and the space they take up on the accelerator's architecture).
Pazos, a member of the KAUS team, shared details about the fabrication process, stating, "A memristor was constructed using an innovative two-dimensional layered material called hexagonal boron nitride. Employing a scalable and cost-effective inkjet printing technology, we printed silver electrodes onto this material." Pazos emphasized, "What makes this achievement truly remarkable is that the unique properties of the 2D h-BN material persist even after the electrode printing process, resulting in enhanced power capabilities and the generation of superior random signals."
The finished TRNG generator, which generated 7 million random bits per second using memristor technology, appeared to meet the team's expectations: it demonstrated the best TRNG performance in terms of stability of its random signal over time, incredibly low energy consumption, and finally, easy and quick circuit readout.
In addition, Pazos highlighted that a functional circuit was demonstrated, featuring the interconnection of the memristor with a commercial microcontroller. This integration enabled live experiments for on-the-fly generation of random numbers.
In contrast to most other technological advances, it also seems as though this technology is ready for use right now. Applications for the Internet of Things (IoT) and other edge devices, including sensor node arrays, might easily adopt the technology.