ReRAM-Based Pseudo-True Random Number Generator With High Throughput and Unpredictability Characteristics

Hardware implementation of a true random number generator integrating a hexagonal boron nitride memristor with a commercial microcontroller

“…The current stochastically fluctuates between two levels (i.e., L 1 ~50 nA and L 2 ~90 nA), and the devices reach optimal functioning for >1 hour at 70 mV (see Figure 2a-c and Supplementary Figure 3). This type of RTN current signals were previously observed in metal-oxides, and they were attributed to charge trapping and de-trapping at atomic defects, normally an oxygen vacancy 38 or an oxygen interstitial ion 39 . In our devices, this behaviour is related to the presence of native defects within the volume of the h-BN stack, mainly lattice distortions formed during the CVD process at h-BN grain boundaries and on substrate imperfections, resulting in amorphous, boron vacancy-rich regions 41 .…”

“…3 †). These types of RTN current signals were previously observed in metal oxides, and they were attributed to charge trapping and de-trapping at atomic defects, normally an oxygen vacancy 38 or an oxygen…”

“…The advantage of this approach is that the resistance change is related to electron trapping and de-trapping at the defects of the insulator, which reduces the energy consumption and slows down the degradation of the MIM nanocell. However, some works only report circuit level simulations driven by pre-recorded RTN acquisitions 19,38,39 ; some experimental demonstrations require a memory array of devices 33,34,36,38 and/or complex hardware 36 or laboratory characterization equipment 35 to prove their functionality; and some TRNG alternatives require computationally-expensive signal post-processing to achieve sufficient output randomness 34 . More importantly, the two main reliability challenges of TRNG circuits based on RTN harvesting are not addressed in any of those works; such challenges are: i) the correct functioning of the TRNG even despite the unstable nature of the RTN effect, which may vanish and reappear randomly through time or disappear altogether due to progressive oxide degradation 40 , and ii) the low resistance ratio between the two RTN levels, which difficult state identification 41 .…”

“…27, 28 and 30 went farther and implemented parts of the circuit with components-of-the-shelf mounted on a protoboard, but the throughput was only 6 kilobit per s, 30 the power overhead of the entropy source was too high in the low resistance state, 31 and it required very high operating voltages. [26][27][28][29][30][31][32][33][34][35][36][37][38][39] proposed that the random telegraph noise (RTN) current signals produced by memristors (i.e., stochastic current fluctuations between two or more levels when a low and constant voltage is applied) could be used as an entropy source in TRNG circuits. The advantage of this approach is that the resis-tance change is related to electron trapping and de-trapping at the defects of the insulator, which reduces the energy consumption and slows down the degradation of the MIM nanocell.…”

Cryptographically strong random number generation using integrated CMOS photodiodes for low-cost microcontroller based applications

“…If the seed and the algorithm are known, the random number can be guessed. However, their fast execution time makes them very suitable for most digital systems [3][4][5][6][7], and therefore they are very popular in most embedded system applications. For example, the highly practiced and standardized C programming language contains a pseudo-random number function called rand() in the stdlib library.…”