This paper describes a full-entropy 128-b key generation platform based on a 1024-b hybrid physically unclonable function (PUF) array, fabricated in 14-nm trigate high-k/metal-gate CMOS. Delay-hardened hybrid PUF cells use differential clock delay insertion to favor circuit evaluation in the desired direction while leveraging burn-in-induced aging for selective bit destabilization enabling quick identification and masking of unstable cells, and subsequent temporal-majority-voting with soft dark-bit masking to reduce PUF bit error by 3.9 times to 1.45% resulting in ~5 ppb failure probability. A stable full-entropy 128-b key is finally generated from the 1024 raw PUF bits using BCH error correction and AES-CBC-based entropy extraction. An all-digital design with compact PUF cell layout occupying 1.84 μm2 achieves: 1) 4-fJ/b energy-efficiency with 3-μW leakage at 0.65 V, 70°C; 2) peak operating frequency of 1 GHz resulting in 1.2-μs key generation latency; 3) robust operation with stable key generation across 0.55-0.75 V, and 25°C-110°C; 4) 14 times separation between intra/inter-PUF hamming distances with 0.99993 entropy ensuring cryptographic quality randomness and uniqueness; 5) 48% higher PUF stability with long-term aging by leveraging transistor degradation to reinforce favorable cell bias; and 6) resiliency to power cycling attacks with common centroid clock routing measured from 49.5% hamming distance between array's evaluation and wake-up states.
3 January 2019 in IEEE Journal of Solid-State Circuits
Cryptographic circuits such as advanced encryption standard (AES) are vulnerable to correlation power analysis (CPA) side-channel attacks (SCAs), where an adversary monitors chip supply current signatures or electromagnetic (EM) emissions to decipher the value of embedded keys. This article describes an all-digital, fully synthesizable SCA-resistant 16-b serial AES-128 hardware accelerator fabricated in 14-nm CMOS, occupying 4900 μm². Randomized byte-order shuffling through heterogeneous Sboxes, linear masked MixColumns, and dual-rail AddRoundKey circuits enable: 1) 9.2x lower correlation between current signatures and hamming distance (HD)/hamming weight (HW) power models compared to an unprotected AES implemented in 14-nm CMOS; 2) 2.3x attenuation of a correlation ratio for correct key guesses; 3) 839-Mb/s encryption throughput with 11-mW total power consumption measured at 750 mV, 25 °C; 4) peak energy efficiency of 390 Gbps/W measured at an energy optimal point of 290 mV, 25 °C, representing an overhead of 23% over the unprotected AES engine; 5) 1200x improvement in minimum-traces-to-disclosure (MTD) over an unprotected AES accelerator, with no successful CPA attacks observed after 12M encryptions; and 7) >1100x improvement in test vector leakage assessment (TVLA ) metric in power and EM time- and frequency-domain analyses.
28 January 2021 in IEEE Journal of Solid-State Circuits
A side-channel attack (SCA) hardened AES-128 and RSA crypto-processor in 14-nm CMOS with measured resistance to correlation power/electromagnetic analysis (CPA/CEMA) in both time and frequency domains is demonstrated. While previously reported linear low-dropout regulators (LDOs) offer improvements in minimum-time-to-disclose (MTD) of extracted key bytes in the time domain, their transformations are less effective against frequency-domain attacks. This article describes a non-linear digital LDO (NL-DLDO) with control loop randomizations that bolster SCA resistance in the frequency domain. The NL-DLDO cascaded with an AES engine augmented with arithmetic countermeasures enables >250Kx improvement in MTD, with no CPA/CEMA/DNN attacks detected after 1-B encryptions, with 8% power and 10% area overheads incurred by arithmetic techniques. The RSA-4K crypto-processor implements exponent magnitude and timing randomizations along with dynamic memory addressing to mitigate time- and frequency-domain attacks. The countermeasures enable 711x suppression in means separation in current/EM magnitudes from 3.1 mV to 4.35 μV, reducing attacker's accuracy to an ineffective random guess classification, while limiting area and performance overheads to <0.05% and 3.25%, respectively.
This article consists of a collection of slides from the author's conference presentation on Swizzle Switch networks for use in many-core systems. Some of the specific topics discussed include: the special features and design of swizzle-switch networks; data routing capabilities; system architecture; processing capabilities; interconnects; and network evaluation techniques.