CAREER: Security Applications of DRAM Cell Decay Effects

该研究项目开发的想法和设计可用于指纹或认证计算机硬件，生成加密密钥，或形成新的安全协议的基础，所有这些都利用了计算机硬件的物理特性。特别是，这项研究的重点是在今天的计算机中发现的商品存储器，这项工作促进了对存储器的物理特性如何用于计算机安全的基本理解。通过关注商品计算设备中已经存在的硬件，本研究中开发的思想可以很容易地部署。通过开展专业和教育活动，以及与国际同行的互动，该项目还支持向广大受众传播研究思想。为了推进硬件安全研究领域，本研究项目探索动态随机存取存储器（dram）中的细胞衰减效应，并将其应用于基于dram的物理不可克隆功能（puf）的设计，以及基于商品dram的新的虚拟现实证明和物理加密原语的开发。通过手动刷新方法的设计，本研究表明，即使DRAM是商品计算机系统的一部分并被软件积极使用，也可以测量DRAM的性能。还评估了环境对DRAM的影响，并加深了对环境如何影响基于DRAM的安全原语的理解。此外，本研究还解决了诸如开发用于DRAM PUF的纠错方案的理论和数学问题，或加速DRAM PUF读出时间的实际问题等挑战。

This research project develops ideas and designs that can be used to fingerprint or authenticate computer hardware, to generate cryptographic keys, or form basis of new security protocols, all leveraging the physical properties of the computer hardware. Especially, the focus of this research is on commodity memories found in today's computers, and the work advances fundamental understanding of how the physical properties of memories can be used in computer security. By focusing on hardware already present in commodity computing devices, the ideas developed in this research can be readily deployed. Through development of professional and educational activities, and interaction with international colleagues, this project also supports dissemination of research ideas to a broad audience.To advance the hardware security research field, this research project explores the cell decay effects in Dynamic Random Access Memories (DRAMs) and applies them to the design of DRAM-based Physically Unclonable Functions (PUFs), as well as to development of new virtual proofs of reality and Physical Cryptography primitives, all based on commodity DRAMs. Through design of a manual refresh approach, this research shows how the DRAM's properties can be measured even if DRAM is part of a commodity computer system and actively used by software. Environmental impacts on DRAM are also evaluated and understanding is developed of how the environment affects DRAM-based security primitives. In addition, this research tackles challenges such as theoretical and mathematical issues of developing error correction schemes for use with DRAM PUFs, or practical issues of accelerating DRAM PUF readout time.