190
S. Hu et al.
In the past ten years, RF fingerprint extraction and identification technology [2] of wireless communication equipment have received extensive attention
all over the world. This method extracts the ‘RF fingerprint’ of the device by
analyzing the communication signal of the wireless device. Due to the hardware
characteristics, different wireless devices have their own unique RF fingerprints,
which can be extracted by analyzing the received RF signals. The method of
extracting hardware features of devices based on communication signals is called
‘radio frequency fingerprint extraction,’ and the method of identifying different
wireless devices using radio frequency fingerprints is called ‘radio frequency identification.’
In the field of correctly identifying communication devices and increasing
wireless communication security, scholars have done a lot of researches on ‘radio
frequency fingerprint extraction.’ Most achieve RF identification by analyzing
the transient signals, such as extracting the radio frequency fingerprint in the
Bluetooth communication signal [3]. However, this technology requires too much
precision of identification device. Recently, RF fingerprint extraction based on
steady-state signals and identification technology has received widespread attention [4,5].
This paper proposes a classification method by integrating different RF fingerprint features and unique classifiers and carries out extensive experiments to
evaluate the performance. In particular, we employ six features, including two
fractal dimension features, phase noise spectrum, and three constellation features extracted from the received information. Three classifiers are adopted to
adaptively combine different features. A test bed is constructed by a low-cost
USRP SDR platform as transceivers. During our work, extensive experiments to
investigate the classification performance under different channel conditions are
carried out. And experimental results show that the proposed RF fingerprinting identification method has an excellent classification performance even at low
SNR conditions.
The content of this paper is arranged as follows: Sect. 2 introduces the extraction and recognition model of RF fingerprinting. Section 3 explains the main
fingerprint features extracted in the experiment and their physical meanings.
In Sect. 4, we describe the construction of the RF fingerprinting experimental environment and the process of data processing. Section 5 shows the main
recognition results for different classifiers and different identification parameters.
Finally, this paper is summarized in Sect. 6.
2 Overview of RF Fingerprinting Model
The RF fingerprinting extraction and identification process of wireless communication device [6] are mainly composed of data generation and acquisition module,
as well as data, preprocess, and classification. As shown in Fig. 1, during data
generation and acquisition module, after the radio frequency signal is sent by
the wireless transmitter, the received signal is sent to the identification system
through some necessary process. In Fig. 2, during the data classification process,
S. Hu et al.
In the past ten years, RF fingerprint extraction and identification technology [2] of wireless communication equipment have received extensive attention
all over the world. This method extracts the ‘RF fingerprint’ of the device by
analyzing the communication signal of the wireless device. Due to the hardware
characteristics, different wireless devices have their own unique RF fingerprints,
which can be extracted by analyzing the received RF signals. The method of
extracting hardware features of devices based on communication signals is called
‘radio frequency fingerprint extraction,’ and the method of identifying different
wireless devices using radio frequency fingerprints is called ‘radio frequency identification.’
In the field of correctly identifying communication devices and increasing
wireless communication security, scholars have done a lot of researches on ‘radio
frequency fingerprint extraction.’ Most achieve RF identification by analyzing
the transient signals, such as extracting the radio frequency fingerprint in the
Bluetooth communication signal [3]. However, this technology requires too much
precision of identification device. Recently, RF fingerprint extraction based on
steady-state signals and identification technology has received widespread attention [4,5].
This paper proposes a classification method by integrating different RF fingerprint features and unique classifiers and carries out extensive experiments to
evaluate the performance. In particular, we employ six features, including two
fractal dimension features, phase noise spectrum, and three constellation features extracted from the received information. Three classifiers are adopted to
adaptively combine different features. A test bed is constructed by a low-cost
USRP SDR platform as transceivers. During our work, extensive experiments to
investigate the classification performance under different channel conditions are
carried out. And experimental results show that the proposed RF fingerprinting identification method has an excellent classification performance even at low
SNR conditions.
The content of this paper is arranged as follows: Sect. 2 introduces the extraction and recognition model of RF fingerprinting. Section 3 explains the main
fingerprint features extracted in the experiment and their physical meanings.
In Sect. 4, we describe the construction of the RF fingerprinting experimental environment and the process of data processing. Section 5 shows the main
recognition results for different classifiers and different identification parameters.
Finally, this paper is summarized in Sect. 6.
2 Overview of RF Fingerprinting Model
The RF fingerprinting extraction and identification process of wireless communication device [6] are mainly composed of data generation and acquisition module,
as well as data, preprocess, and classification. As shown in Fig. 1, during data
generation and acquisition module, after the radio frequency signal is sent by
the wireless transmitter, the received signal is sent to the identification system
through some necessary process. In Fig. 2, during the data classification process,