| This work addresses various Open Systems Interconnection (OSI) Physical (PHY) layer mechanisms to extract and exploit RF waveform features ("fingerprints") that are inherently unique to specific devices and that may be used to provide hardware specific identification (manufacturer, model, and/or serial number). This is addressed by applying a Dual-Tree C omplex Wavelet Transform (DT- C WT) to improve burst detection and RF fingerprint classification. A "Denoised VT" technique is introduced to improve performance at lower SNRs, with denoising implemented using a DT- C WT decomposition prior to Traditional VT processing. A newly developed Wavelet Domain (WD) fingerprinting technique is presented using statistical WD fingerprints with Multiple Discriminant Analysis/Maximum Likelihood (MDA/ML) classification. The statistical fingerprint features are extracted from coefficients of a DT- C WT decomposition. Relative to previous Time Domain (TD) results, the enhanced WD statistical features provide improved device classification performance. Additional performance sensitivity results are presented to demonstrate WD fingerprinting robustness for variation in burst location error, MDA/ML training and classification SNRs, and MDA/ML training and classification signal types. For all cases considered, the WD technique proved to be more robust and exhibited less sensitivity when compared with the TD Technique. |
