Date of Completion

3-26-2019

Embargo Period

3-25-2020

Keywords

phase change memory, hardware security, physical unclonable function, device physics

Major Advisor

Professor Helena Silva

Associate Advisor

Professor Ali Gokirmak

Associate Advisor

Professor John Chandy

Associate Advisor

Professor Rajeev Bansal

Field of Study

Electrical Engineering

Degree

Doctor of Philosophy

Open Access

Open Access

Abstract

Security has become a crucial concern in hardware design due to the growing need of protection in everyday financial transactions and exchanges of private information. Physical unclonable functions (PUFs) utilize the inevitable process variations and other stochastic properties in devices to provide a unique way to verify trusted users during access to hardware devices. Improvements in attack methods have recently moved the field of PUFs from traditional silicon devices toward emerging non-volatile memories, such as phase change memory (PCM). Due to intrinsic cell-to-cell and cycle-to-cycle programming variability and high endurance properties of PCM nano-devices, unpredictable and reconfigurable PUF challenge-response pairs can be achieved. This programming variability, which comes in addition to the process variations present in any technology, is an important advantage of PCM for implementations of PUFs and other hardware security primitives.

In this work, programming variability in PCM nano-devices are electrically characterized using various cell dimensions and pulsing techniques for PUF applications. The underlying contributing factors, originating from external circuitry and phase change dynamics of PCM nano-devices, that enhance programming variability are identified by performing post-measurement scanning electron microscopy (SEM) imaging, further electrical characterization, SPICE modeling of the experimental setup, and finite element simulations of PCM devices.

Once programmed, the short and long-term stability of the programmed states in PCM nano-devices is monitored, and the spontaneous resistance evolution trends for different cell types are studied, which helped in identifying the cell geometries that result in long data retention time for memory devices, and those that result in earlier data loss creating opportunities for new security applications, such as time-sensitive memory devices. The disturbance to the spontaneous resistance evolution at any programmed state caused by SEM imaging is also characterized. Imaging is observed to leave remarkable evidence of tampering posing resistance toward reverse engineering and ensuring robust security to the PCM-based nano-devices. Lastly, the reconfigurable source of randomness in PCM-based PUFs relying on programming variability is contrasted with a static source of randomness in a ZnO nanoforest-based PUF. The advantageous reconfigurability in PCM-based PUFs can ensure resistance toward physical attacks, which is not achieved with the ZnO nanoforest-based PUF.

COinS