Introduction of Random Forest Classifier to ZigBee Device Network Authentication Using RF-DNA Fingerprinting

摘要

The decentralized architecture of ZigBee ad-hoc networks creates unique security challenges to ensure only authentic devices are granted network access. Non-parametric Random Forest (RndF) and Multi-Class AdaBoost (MCA) ensemble classifiers were introduced with RF-Distinct Native Attribute (RF-DNA) fingerprinting to enhance device authentication performance. Correct classification (%C) performance is improved up to 24% over other classifiers, with 10% improvement at the lowest SNR = 0.0 dB. Network intrusion tests correctly rejected 31/36 rogue devices vs. 25/36 and 28/36 with previously used classifiers. The key benefit of ensemble method processing is improved rogue rejection in noisy environments-gains of up to G_s = 18.0 dB are realized over other classifiers. Collectively considering demonstrated %C and rogue rejection capability, the use of ensemble methods improves ZigBee network authentication and enhances anti-spoofing protection afforded by RF-DNA fingerprinting.