PLC Hardware Discrimination using RF-DNA fingerprinting

摘要

Programmable Logic Controllers are used to control and monitor automated process in many Supervisory Control and Data Acquisition (SCADA) critical applications. As with virtually all electronic devices, PLCs contain Integrated Circuits (IC) that are often manufactured overseas. ICs that have been unknowingly altered (counterfeited, manufactured with hardware Trojans, etc.) pose a significant security vulnerability. To mitigate this risk, the RF- Distinct Native Attribute (RF-DNA) fingerprinting process is applied to PLC hardware devices to augment bit-level security. RF-DNA fingerprints are generated using two independent signal collection platforms. Two different classifiers are applied for device classification. A verification process is implemented for analysis of Authorized Device Identification and Rogue Device Rejection. Fingerprint feature dimensional reduction is evaluated both Qualitatively and Quantitatively to enhance experimental-to-operational transition potential. The findings of this research are that the higher quality signal collection platform had a classification performance gain of approximately 10dB SNR. Performance of the classifiers varied between signal collection platforms, and also with the application of fingerprint dimensional reduction. The lower quality signal collection platform saw a maximum gain of 5dB SNR using reduced dimensional feature sets compared against the full dimensional feature set.