IEEE Circuits and Systems Magazine - Q3 2021 - 13

Security of CMOS based TRNGs become one of the major concerns.
Additional circuitry is required to counter security attacks on TRNGs.
low-cost and low-power TRNG is proposed wherein it
achieved relatively lower bitrate [27]. In order to avoid
the PFD, an even stage RO based TRNG is demonstrated
that explore oscillator collapse and achieves lower
energy consumption [28]. Regardless of lower energy
consumption, in order to obtain sufficient oscillations
to collapse, this TRNG requires a tuning algorithm that
needs to run on an off-chip host processor. Another
TRNG concept is introduced in the literature using the
metastability of digital circuits [80]. High-speed metastability
based TRNG is proposed that utilizes comparator
in common-mode [81]. Three independent selfcalibrating
random sources based all-digital TRNG is
demonstrated that achieved a full entropy [30]. Regardless
of the advantages, CMOS based TRNGs show many
challenges that are listed as follows:
■ Although CMOS based circuits are promising for
general-purpose digital, analog, and mixed-signal
applications, they offer limited security specific
properties for designing TRNG.
■ CMOS based TRNGs relatively exhibit poor randomness.
Consequently, require large post-processing
units, which increases the area and energy
consumption.
■ Security of CMOS based TRNGs become one of the
major concerns. Additional circuitry is required
to counter security attacks on TRNGs[16].
■ The complexity involved in CMOS based TRNG
design often impractical for many emerging applications
that rely on resource-constrained Internet-of-Things.
Several
TRNG designs based on beyond-CMOS device
technologies exhibit smaller area and lower energy
consumption compared to CMOS[82]. The researcher's
investigation recently revealed that the self-heating
phase-change RAMs exhibit larger threshold dispersion
compared to other non-volatile memory concepts
[83]. Consequently, it is demonstrated that the phasechange
RAMs are highly suitable to design simple and
promising TRNG. Carbon nanotube FETs exhibit high
entropy by exploiting chirality, placement, spacing, and
dimensions, and other physical variations of channel
nanotube [83]. However, the extraction of digital random
bits from these inherent physical phenomena is
proved to be challenging. A new group of generators explores
stochastic behavior in nanoscale memristors like
RRAMs and STT-MTJs. High-speed and energy-efficient
THIRD QUARTER 2021
TRNG designs have been introduced that explore intradevice
stochastic variations and RTN in resistive RAM
[84]-[86]. Among different memristor-based TRNGs,
STT-MTJ based TRNGs have attracted wide attention
due to low power consumption and higher generation
speed [74], [88]. However, TRNG based on a single STTMTJ
device exhibited higher bias in the output sequence
due to the fabrication variations. To obtain an acceptable
level of randomness, a variation resilient multiple
STT-MTJ based TRNG is proposed that demonstrates a
lower energy consumption [54]. A new method of TRNG
design is demonstrated using the program disturb and
read noise characteristics of flash memory cells [89][90].
The TRNG using flash memory exhibited high tolerance
against aging and temperature effects. Recently,
III-V InAs TFET based TRNG is also proposed that leverages
the p-i-n forward leakage of the TFET [60]. It is
demonstrated that by exploiting the TFET unique p-i-n
forward leakage, the TRNG is resulted to be energy efficient
and compact.
Figure 11(a) presents the comparison of bit-energy
consumption against the area of emerging TRNG designs
with CMOS based designs.
It can be observed
that among various beyond-CMOS TRNG designs demonstrated,
STT-MTJ and TFET based TRNGs achieve
lower energy consumption. On the other hand, CMOS
based TRNG designs have shown degraded performance
compared to emerging technologies based
TRNG designs. Similarly, Fig. 11(b) presents TRNGs bitrate
against the area wherein STT-MTJ shows a higher
bit rate compared to the other designs. Additionally,
Fig. 11(c) presents bit-energy consumption against bitrate
and demonstrated that the STT-MTJ based TRNGs
have shown lower energy consumption with higher
speed compared to other designs. From this analysis,
it can be concluded that the STT-MTJ based TRNG designs
exhibited superior performance compared to the
other designs.
ii. Physically Unclonable Functions
Physically unclonable functions have emerged as secure
on-the-fly key generators that utilize inherent physical
variations [91]. PUFs avoid the storing of keys in volatile/
non-volatile memory and provide volatile tamper resistant
solutions to the key storage. However, PUF suffers
from several problems like environmental variations, aging,
and electromagnetic interference. Due to this PUF
IEEE CIRCUITS AND SYSTEMS MAGAZINE
13
IEEE Circuits and Systems Magazine - Q3 2021

Table of Contents for the Digital Edition of IEEE Circuits and Systems Magazine - Q3 2021
