SIMULATION AND PERFORMANCE ANALYSIS OF MEMRISTOR BASED RING OSCILLATOR TRUE RANDOM NUMBER GENERATOR FOR HARDWARE SECURITY

• Main Author: NOOR ALIA BINTI NOR HASHIM
• Format: text
• Language: English
• Published: 2023
• Subjects: SIMULATION AND PERFORMANCE ANALYSIS OF MEMRISTOR

The world is evolving where there is a huge reliance on the internet and online transactions which makes hardware security exposed to security attacks. Software operations are easily targeted which makes it very important to update the hardware security. It is crucial to investigate all new and improved solutions in particular using memristors. This research presented a memristor-based ring oscillators TRNG (MROTRNG) using CMOS technology. The memristor possesses characteristics such as memory-like behaviour, relatively small in size and is compatible to be used in CMOS fabrication. These characteristics are suitable to be implemented in the true random number generator (TRNG) design and be applied for hardware security purposes. By implementing memristor in the TRNG design of the ring oscillators, the oscillation frequency can be more random and affects the process variations of the entropy source circuit. The memristor based ring oscillators was designed to oscillate around 100MHz with a voltage supply of 1.8V using SilTerra‟s 180nm CMOS technology. The randomness value was calculated to be 1.6769 for MRO. Performance metric were calculated and analyzed in terms of uniqueness, uniformity, and bit-aliasing with an ideal value of 50% for each metric. In addition, the National Institute Statistical Test Suite (NIST) was used to evaluate the performance results of the TRNG. There are three scenarios that were used to compare the performance of the TRNG designs. These are 1 st scenario: TRNG with inverters of NMOS and PMOS transistors, 2nd scenario: TRNG with common source with resistive load as inverters and 3rd scenario: TRNG with common source with memristive load as inverters. Each scenario of the design passes the performance metric calculation but the best results was based on 2nd scenario with the smallest percentage error of 0.208% with 49.986%, 49.03% and 49.03% of uniqueness, uniformity and bit-aliasing respectively when data is 10000 bits in size. Although, the TRNG design in 2nd scenario produces the best results and passed 10 out of 12 of the NIST tests compared to other scenarios, the design in 3rd scenario showed that the output also demonstrates randomness in the output. The proposed memristor based ring oscillator TRNG was implemented with four different memristor window functions to observe any changes that might occur in the output bits. All of the TRNG design in different memristor window function passed 10 out of 12 of the NIST tests when data size is 2000 bits. This research has proved that the memristor based ring oscillator TRNG has the potential to produce output that demonstrates randomness characteristics suitable as solution to security.