Project Details
Projekt
Intrinsic Physical Unclonable Functions from Emerging Non-Volatile Memories
Subject Area
Security and Dependability, Operating-, Communication- and Distributed Systems
Term
from 2020 to 2024
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 440182124
Recent developments have led to the proliferation of resource constrained devices that undertake increasingly extensive and decisive tasks. Based on physical process variations during semiconductor manufacturing, memory-based physical unclonable functions (PUFs) provide lightweight, cost-efficient, and flexible hardware-based security primitives to protect these devices. Concurrently with these trends, the integration density of circuits is approaching the so-called scaling limit, where structures become so small that the saved information are hard to retrieve and that consequently brings novel non-volatile memory (NVM) types to the scene.In this project, we investigate the realization of intrinsic PUFs on commercially available NVMs in the form of magnetoresistive (MRAM), resistive (ReRAM) and ferroelectric (FRAM) random access memory.For MRAM, a relatively high number of approaches to implement extrinsic PUFs showcased on custom memory cells exists in related literature. Hence to obtain new findings, we will initially refine existing methods to realize intrinsic PUF instances on off-the-shelf MRAM. Methods for realizing PUFs on ReRAM, so far, are covered to a lesser extent in related works. Therefore at first, we will perform experiments on custom-tailored ReRAM memory cells. In this way, we are able to vary different physical properties of memory cells and can directly test their impact on PUF behavior. In the next step, we will use previously gained insights and transfer successful strategies to off-the-shelf ReRAM modules. Finally, we will focus on FRAMs, for which no generally known works about operational PUF instances have been publicized.Therefore, we will use the experience gained over the course of the project to find suitable ways to excite characteristic behavior on FRAM.All obtained PUF instances will be systematically characterized according to established quality metrics.As PUFs from conventional memories are susceptible to varying temperature and supply voltage as well as magnetic fields and radiation, different environmental conditions for the characterization of PUF instances on NVMs will be created. In the next step, we will quantify the results and apply advanced techniques such as protocols, error correcting codes, stochastic models etc. to improve PUF quality and resilience to influences from the environment.In particular when used as random access memory in present-day computers, credentials and results of cryptographic operations are directly accessible on NVMs due to their inherent storage properties.We plan to overcome this drawback by employing self-encryption, where the same NVM is used for 2 different purposes: as a memory to store data and as a PUF for retrieving the key to encrypt this data.
DFG Programme
Priority Programmes
