A Magnetic Tunnel Junction Based True Random Number Generator with Conditional Perturb and Real-Time Output Probability Tracking Won Ho Choi*, Yang Lv*, Jongyeon Kim, Abhishek Deshpande, Gyuseong Kang, Jian-Ping Wang, and Chris H. Kim *equal contribution University of Minnesota, Minneapolis 1
Outline of Presentation • True Random Number Generator (TRNG) • Magnetic Tunnel Junction (MTJ) • MTJ-based TRNG • Conditional perturb scheme • Real-time output probability tracking • Conclusions 2
An Application of True Random Number Generator (TRNG) Q. Tang, et. al., CICC, 2014 • Generates independent, unpredictable, nondeterministic, and aperiodic random numbers • Use random numbers to generate secret keys 3
Prior Art of Physical TRNG • Direct noise amplification from devices – Random Telegraph Noise ( R. Brederlow, ISSCC, 2006 ) – Resistor thermal noise ( V. Kaenel, CICC 2007 ) – Requires post-processing to achieve sufficient randomness • ROSC based TRNG (M. Bucci, Tran. on Comp., 2003; Q. Tang, CICC, 2014) – Harvesting noise from oscillator jitter – Generally requires noise amplification otherwise yield with low efficiency, thus increases design complexity • Metastability TRNG (C. Tokunaga, JSSC, 2008; S. Mathew, JSSC, 2012) – Inverter pair driven to metastable state – Requires continuous calibrating loop 4
Magnetic Tunnel Junction (MTJ) • Spin polarized electrons rotate the magnetization direction of free layer with spin torque 5
Switching Probability of an MTJ H. Zhao, et. al., JAP, 2011 • Random thermal fluctuation in an MTJ can be utilized for generating random bits • Trade-off relationship between speed, switching energy, and reliability • Switching probability is sensitive to operating conditions 6
MTJ-Based TRNG - Unconditional Reset Scheme - S. Yuasa, et. al., IEDM, 2013, concept only • Applies large reset voltage in every cycles thereby, adversely effecting on TRNG performance 7
Proposed Conditional Perturb Scheme • Perturbs the MTJ according to the previously sampled MTJ state, thereby eliminating the reset phase 8
MTJ Time-to-Breakdown Analysis Failure (%) C. Yoshida, et al., IRPS, 2009 • Absence of a reset phase enhances the lifetime of the MTJ 9
Fabricated MTJ Device • Fabricated MTJ device is used for demonstration of the MTJ-based TRNG 10
Measurement Setup • Random number generator measurement setup with sub-50 picosecond pulse width resolution. 11
Measured Probability Unconditional reset scheme Conditional perturb scheme • A small number of segments fail to meet 50 ± ± ± ± 1% probability 12
Measured Randomness Unconditional reset scheme Conditional perturb scheme # of segments: 55 Test Pass/Fail Frequency Fail 1 Block frequency Pass 2 Cumulative Sums 3 Fail Runs Pass 4 Pass 5 Longest-Run-of-Ones Rank Pass 6 FFT Pass 7 Non-overlapping 8 Pass Template Matching Serial Pass 9 10 Approximate Entropy Pass • Both schemes show a similar level of randomness • The output data fail to pass the frequency and cumulative sums tests 13
Real-Time Output Probability Tracking • Simple single-parameter feedback control • The proposed techniques were implemented in LabVIEW TM and experimentally verified using a fabricated MTJ device 14
Measured Probability and Randomness - Real-Time Output Probability Tracking- Raw data after probability tracking Conditional perturb scheme, # of segments: 55 Test Pass/Fail Frequency Pass 1 Block frequency 2 Pass 3 Cumulative Sums Pass Pass 4 Runs Pass 5 Longest-Run-of-Ones Rank Pass 6 FFT Pass 7 Non-overlapping 8 Pass Template Matching Pass Serial 9 Approximate Entropy Pass 10 • Proposed conditional perturb and real-time probability tracking achieves a good randomness while improving the reliability, speed, and power 15
TRNG Performance Comparison *S. Yuasa, et. al., IEDM, 2013 • Conditional perturb scheme improves the speed, switching energy and reliability 16
A Possible Application with STT-MRAM • It could potentially allow massive generation of random numbers with negligible circuit overhead 17
Conclusions • We demonstrate for the first time a True Random Number Generator (TRNG) based on the random switching probability of Magnetic Tunnel Junctions (MTJs) • Proposed conditional perturb and real-time output probability tracking achieves a good randomness while improving the reliability, speed, and power 18
Recommend
More recommend
