Using Spectral Fingerprints to Improve Wireless Network Security

The proliferation of affordable RF communication devices has given every individual the capability to communicate voice and/or data worldwide. This has increased wireless user exposure and driven the need for improved security measures. While earlier works have primarily focused on detecting and mitigating spoofing at the MAC layer, there has been a shift toward providing protection at the PHY layer by exploiting RF characteristics that are difficult to mimic. This research investigates the use of RF "fingerprints" for classifying emissions by exploiting transient signal features to provide hardware-specific identification. Reliable transient detection is the most important step in the process and is addressed here using variance trajectory of instantaneous amplitude and instantaneous phase responses. Following transient detection performance characterization, power spectral density fingerprints are extracted and spectral correlation used for classification. For proof-of-concept demonstration, the overall detection and classification process is evaluated using experimentally collected 802.11a OFDM signals. Results show that amplitude-based transient detection is most effective. Classification performance is demonstrated using three devices with overall classification accuracy approaching 80% for 802.11a signals at SNRs greater than 6 dB.