A Machine Learning Attack Resilient True Random Number Generator Based on Stochastic Programming of Atomically Thin Transistors

Abstract

A true random number generator (TRNG) is a critical hardware component that has become increasingly important in the era of Internet of Things (IoT) and mobile computing for ensuring secure communication and authentication schemes. While recent years have seen an upsurge in TRNGs based on nanoscale materials and devices, their resilience against machine learning (ML) attacks remains unexamined. In this article, we demonstrate a ML attack resilient, low-power, and low-cost TRNG by exploiting stochastic programmability of floating gate (FG) field effect transistors (FETs) with atomically thin channel materials. The origin of stochasticity is attributed to the probabilistic nature of charge trapping and detrapping phenomena in the FG. Our TRNG also satisfies other requirements, which include high entropy, uniformity, uniqueness, and unclonability. Furthermore, the generated bit-streams pass NIST randomness tests without any postprocessing. Our findings are important in the context of hardware security for resource constrained IoT edge devices, which are becoming increasingly vulnerable to ML attacks.

All-in-one, bio-inspired, and low-power crypto engines for near-sensor security based on two-dimensional memtransistors

“…This work further advances the field of 2D material-based devices by demonstrating robust security features achievable under the same hardware infrastructure used for sensing, storage, and computing. Note that some of our earlier work has also shown the potential use of 2D materials in resolving rampant security vulnerabilities 57,58,61,62 .…”

“…First, MoS 2 is the most advanced among other 2D materials in terms of scalable growth over a large area (wafer scale) as well as in terms of demonstration of high-performance FETs with on current > 250 µA/µm at ultrashort channel lengths 43 with low device-to-device variability [43][44][45] , which are promising to meet the requirements set forth by the International Roadmap of Devices and Systems (IRDS) 46 . In addition, circuit and architecture level demonstrations of digital, analog, and radio frequency (RF) electronics based on 2D transistors are already available [47][48][49][50][51] , and emerging applications such as neuromorphic, optoelectronic, straintronic, hardware security, and biomimetic technologies exploiting the sensing, computing, and storage capabilities of 2D transistors have also been reported 13,26,28,30,[52][53][54][55][56][57][58][59][60][61] . This work further advances the field of 2D material-based devices by demonstrating robust security features achievable under the same hardware infrastructure used for sensing, storage, and computing.…”

Application of 2D Materials in Hardware Security for Internet‐of‐Things: Progress and Perspective

“…Table 1 lists several key parameters adapted to examine the raw random bits. [16] The randomness of generated keys can be evaluated using the test suites listed in Table 2. [51,52] Conventional practical demonstrations of TRNG using silicon are based on jitter, metastability of cross-coupled inverters, and discrete-and analog-time chaos.…”

“…[7,8,[12][13][14] In the application of hardware security, 2D materials exhibit tremendous variation ascribing to variation in different intrinsic and extrinsic parameters arising from random defects in the synthesis of materials, material transfer, and device fabrication, giving rise to much more complexed cycle-to-cycle (C2C) or device-to-device (D2D) variation than other materials as listed in Table S1, Supporting Information. [10,[15][16][17][18][19] Therefore, 2D material-based security devices can have multiple high-quality entropy sources without increasing energy consumption and hardware overhead to compensate for low randomness as in the case of silicon-based hardware security.…”

A flexible and stretchable bionic true random number generator

“…In contrast, physical TRNGs exploit some unpredictable or, at least, difficult to predict physical process and use the outputs to produce a bits sequence that can be truly random [12], thus enabling superior reliability for data encryption and other applications, such as cybersecurity, stochastic modeling, lottery, or games of chance [15][16][17]. Up to date, a series of TRNGs based on different physical sources with different working mechanisms has been investigated to generate considerable random numbers in lieu of conventional pseudo random numbers, such as random telegraph noise (RTN) based on memristors [18][19][20][21][22], thin-film transistor [23][24][25], and triboelectric generator [26,27], laser chaos [28][29][30], photonic integrated chip [31], quantum entropy sources [32][33][34][35], bichromatic laser dye [36], crystallization robot [37], DNA synthesis [38], and so forth. However, majority of aforementioned existing TRNG implementations rely on rigid platforms and expensive complicated manufacturing crafts, which cannot compatibly adapt the portable networked devices and systems since emerging wearable technologies typically demand low-cost and mechanically flexible security hardware components.…”