Methods to Assess Sensitivity Limits of Second Order Effects Measurements
Abstract
Sensitivity limits of a specific second order effects(2OE) testing method were studied with respect to the detection of a specific set of circuit modifications using flash-enabled FPGAs as a testbed. The 2OE method tested was an expanded variant of Sandia's Power Spectrum Analysis (PSA) and the circuit modifications to be studied were hardware Trojans of different size and with controlled changes to physical layout on the test FPGA. The focus of this work was data analysis comparing the utility of traditional Hotelling t expn 2 and corresponding p-values as separability metrics to confusion matrix analysis derived from machine learning classifiers based on logistic regression. The goal of the analyses was to optimize data collection for 2OE features that improve separability while clarifying the limitations of the method with respect to identifying the specific risk defined by the part/problem pair. The optimized 2OE methods easily identified physical differences in FPGAs (associated with manufacture changes related to lot dates) and certain hardware Trojan varieties were clearly identified as separable while others proved more difficult for the 2OE features analyzed. This work provides a path for2OE methods optimization and quantification of utility to mitigate specific risks.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 29, 2021
- Accession Number
- AD1137023
Entities
People
- Brendan Foran
- Carl T. Boone
- Dmitry Veksler
- Garret Chan
- Salam Zantout
- Sean C. Stuart
- Vikram Rao
Organizations
- The Aerospace Corporation