Comparison of Radio Frequency Distinct Native Attribute and Matched Filtering Techniques for Device Discrimination and Operation Identification

The research presented here provides a comparison of classification, verification, and computational time for three techniques used to analyze Unintentional Radio- Frequency (RF) Emissions (URE) from semiconductor devices for the purposes of device discrimination and operation identification. URE from ten MSP430F5529 16-bit microcontrollers were analyzed using: 1) RF Distinct Native Attribute (RFDNA) fingerprints paired with Multiple Discriminant Analysis/Maximum Likelihood (MDA/ML) classification, 2) RF-DNA fingerprints paired with Generalized Relevance Learning Vector Quantized-Improved (GRLVQI) classification, and 3) Time Domain (TD) signals paired with matched filtering. These techniques were considered for potential applications to detect counterfeit/Trojan hardware infiltrating supply chains and to defend against cyber attacks by monitoring executed operations of embedded systems in critical Supervisory Control And Data Acquisition (SCADA) networks