Date of Award
Summer 9-12-2023
Degree Name
Doctor of Philosophy (PhD)
Degree Type
Dissertation
Abstract
Traditionally, memory devices store information in a static manner be it charge-based devices like floating-gates of FeFET or any other method for that matter such as spin-based, electrochemical-based, magnetic-based, etc. But what if memory storage became dynamic in nature and you can control the dynamics? My thesis is based on these premises where I investigate the theory and application of Dynamic Analog Memory (DAM). In my thesis, I answer the questions on how to design such a DAM and what optimal characteristics it needs to have for it to demonstrate advantages over traditional memory devices. To this end, my thesis first focuses on a self-powered dynamical system that is implemented by depositing charges on an electronically isolated poly-silicon island where the charge leaks through to a semiconductor substrate. The amount of leakage is synchronized not only across multiple tunneling junctions but also across different dies. The dynamical system can be desynchronized by coupling external signals to it. I designed a synaptic memory element that exploits the desynchronization between two dynamical systems to implement an analog memory. First, I characterize the plasticity and the energy required for updating the analog memory. Then I show tunable memory consolidation properties of these synaptic elements using a benchmark random-pattern experiment. Furthermore, I will show that when Fowler-Nordheim quantum tunneling process is used as the leakage mechanism for the DAM, the synaptic elements can exhibit optimal memory consolidation and task-specific based consolidation which can be used in continual learning scenarios. Finally, I also demonstrate that by exploiting the tradeoff between the memory retention period and the energy required for updating the analog memory information can be stored by expending less than 1pj of energy per bit during the training phase of an artificial neural network, four orders of magnitude improvement than conventional memory. Next, I have implemented a novel class of quantum-secure dynamic encryption key distribution and authentication protocols exploiting the security primitives and synchronization capabilities of these self-powered dynamical systems. The FN-dynamical systems are not only synchronized with each other but also the dynamic profile can be modeled accurately with respect to time. I proposed a key exchange protocol that uses publicly available identical copies of self-powered chipsets where the temporal dynamics on the hardware chipsets are synchronized with its software clone i.e. the analytical model running on a server. I show that the dynamic keys derived from these temporal dynamics meet the National Institute of Standards and Technology (NIST) criteria. I also prove the security of these protocols under a standard model and against different adversarial attacks. I have also investigated the robustness of these protocols using hardware results and propose error-correcting protocols to mitigate noise-related artifacts. Finally, I propose a synchronized pseudo-random number generator (SPRNG) that uses a combination of a fast, low-complexity linear-feedback-shift-register (LFSR) based PRNG and a slow but secure, synchronized seed generator based on self-powered FN-dynamical system. I investigated protocols to periodically and securely generate random bits using the self-powered FN-dynamical system for seeding the LFSR. The time-varying random seeds extend and break the LFSR periodic cycles, thus making it difficult for an attacker to predict the random output or the random seed.
Language
English (en)
Chair
Shantanu Chakrabartty