IEEE Circuits and Systems Magazine - Q3 2021 - 12

magnetic orientation of ferromagnetic materials that
means the MTJ device exhibits low resistance (storing
logic state " 0 " ) when the relative magnetizations of the
ferromagnetic layers are parallel, and exhibits high resistance
(storing logic state " 1 " ) when they are antiparallel
as shown in Fig. 9(b). MTJ changes its state when
the bidirectional current through the device is higher
than the critical current.
Despite promising features, STT-MTJ exhibits thermal
fluctuations of magnetization that causes the switching
time behavior as a statistical distribution. The switching
probability (P) in terms of amplitude (I), duration (t) of
the current pulse can be expressed as (2)
PI t (, )
=- S
`1
()
=
exp
TI T0 exp 1
t j
cmI
Ico
2
(2)
(3)
Based
on equations 2 and 3, the STT-MTJ switching
probability is a function of the current (I) and the
pulse duration (t). Furthermore, the STT-MTJ exhibits
different switching characteristics due to the
variation in device size, magnetic materials thickness,
and the material used [74]. Because of superior
intrinsic stochastic behavior and low power consumption
of STT-MTJ, it is widely explored for several
hardware security primitives/methodologies like
TRNG, PUF, DPA countermeasures, and hardware obfuscation
[54], [75], [76].
III. Hardware Security Methodologies
Exploiting Post-CMOS Devices
This section initially presents the review of CMOS based
hardware security primitives and methodologies. The
challenges of CMOS technology in implementing these
techniques have been discussed. Further, a complete
review of post-CMOS devices based hardware security
primitives and methodologies have been presented
with analysis.
A. Hardware Security Primitives
Emerging device technologies based hardware security
primitives like true random number generators and
physically unclonable functions have been discussed in
this sub-section.
Parallel
Free Layer
Fixed Layer
Tunnel Barrier
Oxide Layer
Oxide Layer
Anti-Parallel
Logic '0'
(a)
(b)
Figure 9. STT-MTJ (a) Device structure (b) Switching behavior.
Logic '1'
i. True Random Number Generators
Random number generators (RNGs) are the on-the-fly
key generators used for cryptographic applications [25].
Due to the deterministic algorithms of PRNG, the generated
security keys are statistically weak [25]. To generate
the statistically independent and secure keys, a true
random number generator (TRNG) is introduced that
harvests entropy from a physical phenomenon like jitter,
metastability, and randomness of various electrical
parameters. The architecture of TRNG is as shown in
Fig. 10. It mainly consists of three major modules which
are entropy source, harvester, and post-processing unit
[18]. The entropy source extracts random signals from
an unpredictable physical source. The extracted random
analog signals can be converted into a digital bitstream
using the harvester module [77]. Generally, the quality
of the digital bitstream obtained doesn't bear statistical
robustness. To enhance the bitstream quality, the postprocessing
method is widely used. Based on the type of
physical source of randomness, TRNGs can be broadly
divided into analog and digital TRNGs [18]. While analog
TRNGs are expensive in nature and limited in availability,
the digital TRNGs are compatible and cost-effective.
In the recent past, various digital TRNGs have been
Physical Source and Extraction
Entropy
Source
Harvester
Raw Bits
PostProcessing
Figure
10. General architecture of TRNG design.
12
IEEE CIRCUITS AND SYSTEMS MAGAZINE
Processed
Bits
demonstrated based on popular CMOS technology. One
basic type of CMOS TRNG design was presented in [16]
that exploits oscillator jitter. A three-edge multimode
CMOS ring oscillator (RO) based TRNG is demonstrated
that uses RO collapse time from third harmonic to
the fundamental frequency [78]. However, this method
requires a phase-frequency detector (PFD) to find the
edge collapse event that creates high nonlinearities. A
differential RO with feedback resistors structure can
be used as TRNG that shows high resilience to process
variations [79]. Due to the on-chip resistors, the
resultant power consumption and area of this TRNG
is relatively high. A tetrahedral ring oscillator based
THIRD QUARTER 2021
IEEE Circuits and Systems Magazine - Q3 2021

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