Design of Low-Cost Memory-Based Security Primitives and Techniques for High-Volume Products
Research Report Highlight
This report investigates the feasibility of generating a TRNG using flash memory. A framework is proposed to generate aging-resilient device IDs and to detect recycled flash memory. A similar approach using SRAMs was also validated in Si.
We have investigated the feasibility of generating TRNG using flash memory. We have proposed a framework to generate aging-resilient device IDs and to detect recycled flash memory. For SRAM based approach, we provided a statistical inference framework to estimate the threshold, which discriminates the new devices from the counterfeit ones. Deliverable 2: The proposed frameworks are validated by the silicon data, which are collected from 14 SRAM chips and 20 flash chips. More than 10 GB data are measured from memory chips. Deliverable 3: We utilized a Spartan-6 based system as the controller of both SRAM and flash. The collected memory data are sent to PC through serial computation, such as UART.
The report is organized as followings: In Section 1, we propose a statistical inference based approach to determine if a SRAM is recycled or not. This approach exploits the high-temperature condition to predict the aging sensitive SRAM cells. These cells present the similar behaves as the real aged SRAM cells. Next, a statistical inference based approach is engaged to generate the scores and thresholds for recycled SRAM determination. In Section 2, we provide a comprehensive framework for generating device ID and detecting the recycled flash memory. Our approach exploits the partially programming effects of the flash memory to reveal the page usage information.