Equipment identity recognition method, electronic equipment and storage medium

Abstract

The invention relates to equipment identity recognition based on radio frequency fingerprints, in particular to an identity recognition method, electronic equipment and a storage medium, and the method comprises the steps: obtaining transmission signals, received by signal receiving equipment under different resource grids, of all sending equipment, and forming a transmission signal set; analyzing amplitude distribution of transmission signals on different frequency components according to the transmission signal set; coarse-grained and fine-grained radio frequency fingerprints are extracted respectively, amplitude distribution of signal frequency components in observation time is obtained, signal power distribution is calculated, and frequencies of different signal powers in each power interval are taken as radio frequency fingerprints; designing a multi-path auto-encoder structure, performing fine-grained extraction on the intensity, distribution and position characteristics of signal energy through an intensity encoder, a distribution encoder and a position encoder, and determining unified radio frequency fingerprint representation resistant to time-frequency resource change; and the method can adapt to a communication scene of signal time-frequency resource change in modern mobile communication.

Description

Technical Field

[0001] The present invention relates to device identity recognition based on radio frequency fingerprint, and in particular to a device identity recognition method, electronic equipment and storage medium. Background Art

[0002] When a device sends an electromagnetic wave signal, it will inevitably introduce the device's radio frequency fingerprint feature. This feature is mainly caused by the power difference of the hardware components inside the device. Since each electronic component has a unique power difference, the physical characteristics of each device are also unique. This physical feature is like the "fingerprint" of the device, with uniqueness and difficult to clone characteristics, so this feature is also called physical fingerprint feature or radio frequency fingerprint feature. With the deepening of research on radio frequency fingerprint related technologies, radio frequency fingerprint features are generally considered to be the only features of wireless devices, so they can be used for the identification and authentication of electromagnetic radiation sources. Specifically, the device identification technology based on radio frequency fingerprint features can accurately distinguish wireless devices that use the same frequency, bandwidth, and modulation mode, and has very good practical value. Therefore, the authentication system based on radio frequency fingerprint features can authenticate the connected wireless devices at the physical signal level. Compared with traditional device identity authentication methods, radio frequency fingerprint technology can effectively resist attacks such as forgery and tampering, and has the characteristics of physical non-cloning.

[0003] Current RF fingerprint research mainly focuses on IoT communication systems, such as ZigBee, WiFi, LoRa, etc. These systems use a fixed time-frequency resource allocation mechanism due to relatively simple service requirements. Therefore, most of the existing RF fingerprint research is carried out under the premise that the signal frequency characteristics are fixed. However, this premise no longer holds true in modern mobile communication systems. In order to adapt to diversified services, high-speed mobile support and spectrum efficiency optimization, the allocation of time-frequency resources in modern mobile communication systems is highly dynamic. In addition, the characterization of RF fingerprints is closely related to signal time-frequency resources, and the existing RF fingerprint technology cannot be directly applied to modern mobile communication systems. A unified RF fingerprint characterization method that is resistant to time-frequency resource changes is urgently needed. Summary of the invention

[0004] The purpose of the present invention is to provide a device identity recognition method, an electronic device and a storage medium to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A device identity recognition method, applied to a signal receiving device, is characterized in that the method comprises:

[0007] Acquire the transmission signals received by the signal receiving devices from the various transmitting devices under different resource grids to form a transmission signal set;

[0008] Analyze the amplitude distribution of the transmission signal at different frequency components according to the transmission signal set;

[0009] Coarse-grained and fine-grained RF fingerprint extraction are performed respectively to determine a unified RF fingerprint representation that is resistant to time-frequency resource changes.

[0010] The step of analyzing the amplitude distribution of the transmission signal at different frequency components according to the transmission signal set comprises:

[0011] Performing pre-processing operations of synchronization, normalization, noise reduction and channel response removal on the transmission signal;

[0012] The pre-processed transmission signal is converted from the time domain to the frequency domain using short-time Fourier transform to determine the amplitude variation model of the signal frequency component within the observation time.

[0013]

[0014] in, represents the transmission signal under a certain resource grid, ω(t-τ) represents the window function, τ is the time position of the window sliding on the signal, f is the frequency variable, e -j2πft It is the core of Fourier transform, which is used to convert signals from time domain to frequency;

[0015] By changing the width of the window function ω(t-τ), the signal intensity and distribution characteristics of the time-frequency analysis results at different resolutions are used as the representation of the amplitude variation pattern of the signal at different frequency components.

[0016] Perform coarse-grained and fine-grained RF fingerprint extraction to determine a unified RF fingerprint representation that is resistant to time-frequency resource changes, including:

[0017] Obtain the amplitude distribution of the signal frequency components within the observation time, calculate the signal power distribution, and use the frequency of occurrence of different signal powers in each power interval as the RF fingerprint;

[0018] Design a multi-channel autoencoder structure to extract the intensity, distribution and position features of signal energy in a fine-grained manner through intensity encoder, distribution encoder and position encoder;

[0019] The valid RF fingerprint uniformity representation is confirmed based on the two extracted RF fingerprint features for device identification.

[0020] Obtain the amplitude distribution of the signal frequency components within the observation time, calculate the signal power distribution, and use the frequency of occurrence of different signal powers in each power interval as the RF fingerprint, including:

[0021] Obtain the amplitude distribution model of the signal frequency component within the observation time Calculate the signal power P:

[0022]

[0023] in, and Respectively represent the real part and imaginary part operations;

[0024] Determine the power histogram H(P) according to the signal power P i ): H(P i )=m i {P|P∈[P i,min ,P i,max ]}; where P i Indicates the set power interval, m i represents the number of power values ​​falling into each power interval, P i,min and P i,max are the minimum and maximum values ​​of the i-th power interval, respectively;

[0025] The H(P i ) as a unified representation of radio frequency fingerprint RFF A .

[0026] Design a multi-channel autoencoder structure to extract the intensity, distribution and position features of signal energy in a fine-grained manner through intensity encoder, distribution encoder and position encoder, including:

[0027] Obtain the amplitude distribution model of the signal frequency component within the observation time As the input of the multi-channel autoencoder, the total loss function L of the multi-channel autoencoder total Designed for:

[0028] L total =α·L intensity +β·L distribution +γ·L position ;

[0029] Among them, L intensity , L distribution and L shape denote the loss functions of the intensity encoder, distribution encoder, and position encoder, respectively; α, β, and γ denote the weights of the intensity encoder, distribution encoder, and position encoder, respectively;

[0030] During the encoder training process, the total loss function L is minimized. totalAs the training goal, by increasing the weights of the intensity and distribution encoders and reducing the weight of the position encoder, the loss functions of the intensity encoder, distribution encoder and position encoder are reduced, and finally the total loss function converges; at this time, the encoding results E of the intensity encoder, distribution encoder and position encoder are extracted respectively intensity 、E distribution and E shape , and determine a unified representation of radio frequency fingerprint RFF B :

[0031] RFF B =[E intensity ,E distribution ,E position ].

[0032] The valid RF fingerprint uniformity characterization is confirmed based on the two extracted RF fingerprint features, including:

[0033] A unified representation of RF fingerprint RFF under power histogram is obtained respectively A A unified representation of radio frequency fingerprints under multi-channel autoencoders RFF B , calculate and obtain the agreed radio frequency fingerprint characterization RFF that is resistant to time-frequency resource changes:

[0034] RFF=[RFF A ,RFF B ].

[0035] The sending device has the capability of dynamically allocating time and frequency resources, and sends a baseband signal of a fixed sequence to the signal receiving device.

[0036] An electronic device, comprising:

[0037] at least one processor;

[0038] and, a memory communicatively coupled to the at least one processor;

[0039] The memory stores instructions that can be executed by at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the above-mentioned identification method.

[0040] A computer-readable storage medium stores a computer program, wherein the computer program implements the above-mentioned identification method when executed by a processor.

[0041] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0042] The present invention introduces a device identification method, an electronic device and a storage medium. After the signal receiving device receives a signal with dynamically changing time-frequency resources, a radio frequency fingerprint uniformity characterization model that is resistant to time-frequency resource changes is constructed by combining signal processing and artificial intelligence methods. The method can adapt to communication scenarios with changing signal time-frequency resources in modern mobile communications, and is of great significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0044] Figure 1 It is a flow chart of a device identity recognition method, an electronic device and a storage medium of the present invention;

[0045] Figure 2 It is a research technology roadmap of a device identification method, an electronic device and a storage medium of the present invention;

[0046] Figure 3 It is a structural diagram of a multi-channel autoencoder of a device identity recognition method, an electronic device and a storage medium of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] See also Figure 1-Figure 3 , the present invention provides a technical solution:

[0049] A device identity recognition method, applied to a signal receiving device, is characterized in that the method comprises:

[0050] Acquire the transmission signals received by the signal receiving devices from the various transmitting devices under different resource grids to form a transmission signal set;

[0051] Analyze the amplitude distribution of the transmission signal at different frequency components according to the transmission signal set;

[0052] Coarse-grained and fine-grained RF fingerprint extraction are performed respectively to determine a unified RF fingerprint representation that is resistant to time-frequency resource changes.

[0053] Analyze the amplitude distribution of the transmission signal at different frequency components based on the transmission signal set, including:

[0054] The transmission signal is subjected to preprocessing operations such as synchronization, normalization, noise reduction, and removal of channel response: the signal receiving device obtains the transmission signal through the wireless channel transmission, and performs preprocessing such as synchronization, normalization, and noise reduction to obtain the signal y(t,f):

[0055] y(t,f)=ξ R (h(ξ T (x n (t,f)))), where ξ R represents the RF fingerprint function of the receiving device, and h represents the channel response;

[0056] Among them, the wireless channel h will have an impact on the RF fingerprint, and the channel response can be eliminated by solutions including but not limited to channel equalization, channel independent spectrum, spatial diversity gain, and channel coherence time. At the same time, in the RF fingerprint recognition scenario, the focus is generally on the scenario where one receiving device recognizes multiple sending devices. Therefore, the RF fingerprint of the receiver ξ R As a constant, we can get a simplified signal.

[0057] The pre-processed transmission signal is converted from the time domain to the frequency domain using short-time Fourier transform to determine the amplitude variation model of the signal frequency component within the observation time.

[0058]

[0059] in, represents the transmission signal under a certain resource grid, ω(t-τ) represents the window function, which is used to localize the signal at time τ τ is the time position of the window sliding on the signal, f is the frequency variable, and e -j2πft It is the core of Fourier transform, which is used to convert signals from time domain to frequency;

[0060] From the frequency domain, the effect of RF fingerprint on the transmitted signal can be summarized as the impact on the amplitude of each frequency component of the signal, including the increase or decrease of the amplitude of the existing frequency components of the signal, as well as the addition of new frequency components. The specific amplitude variation pattern on different frequency components is the RF fingerprint feature. By performing time-frequency analysis on the signal, the amplitude distribution of the frequency components at all observation times and frequency points of the signal is obtained;

[0061] By changing the width of the window function ω(t-τ), the time resolution and frequency resolution of the STFT can be adjusted. The wider the window, the higher the frequency resolution and the lower the time resolution; the narrower the window, the higher the time resolution and the lower the frequency resolution; the signal strength and distribution characteristics of the time-frequency analysis results at different resolutions are used as the representation of the amplitude change pattern of the signal at different frequency components.

[0062] Perform coarse-grained and fine-grained RF fingerprint extraction to determine a unified RF fingerprint representation that is resistant to time-frequency resource changes, including:

[0063] Obtain the amplitude distribution of the signal frequency components within the observation time, calculate the signal power distribution, and use the frequency of occurrence of different signal powers in each power interval as the RF fingerprint;

[0064] Design a multi-channel autoencoder structure to extract the intensity, distribution and position features of signal energy in a fine-grained manner through intensity encoder, distribution encoder and position encoder;

[0065] The valid RF fingerprint uniformity representation is confirmed based on the two extracted RF fingerprint features for device identification.

[0066] Obtain the amplitude distribution of the signal frequency components within the observation time, calculate the signal power distribution, and use the frequency of occurrence of different signal powers in each power interval as the RF fingerprint, including:

[0067] Obtain the amplitude distribution model of the signal frequency component within the observation time Calculate the signal power P:

[0068]

[0069] in, and Respectively represent the real part and imaginary part operations;

[0070] Determine the power histogram H(P) according to the signal power P i ): H(P i )=m i {P|P∈[P i,min ,P i,max ]}; where P i Indicates the set power interval, m i represents the number of power values ​​falling into each power interval, P i,min and P i,max are the minimum and maximum values ​​of the ith power interval, respectively; the power histogram shows the frequency of occurrence of different power levels in the time-frequency analysis results, reflects the distribution of signal energy intensity, and is a unified representation of RF fingerprints;

[0071] The H(P i ) as a unified representation of radio frequency fingerprint RFF A .

[0072] Design a multi-channel autoencoder structure to extract the intensity, distribution and position features of signal energy in a fine-grained manner through intensity encoder, distribution encoder and position encoder, including:

[0073] Obtain the amplitude distribution model of the signal frequency component within the observation time As the input of the multi-channel autoencoder, the total loss function L of the multi-channel autoencoder total Designed for:

[0074] L total =α·L intensity +β·L distribution +γ·L position ;

[0075] Among them, L intensity , L distribution and L shape denote the loss functions of the intensity encoder, distribution encoder, and position encoder, respectively; α, β, and γ denote the weights of the intensity encoder, distribution encoder, and position encoder, respectively;

[0076] Design a multi-channel autoencoder structure, and extract the intensity, distribution and position features of signal energy in a fine-grained manner through intensity encoder, distribution encoder and position encoder; here, the signal time-frequency diagram under different resource grids and different frequency resolutions is used as the input of the multi-channel autoencoder; the intensity encoder in the multi-channel autoencoder focuses on extracting features related to the signal energy intensity in the time-frequency diagram; the distribution encoder extracts the distribution features of the signal energy area in the time-frequency diagram; for the position encoder, it is mainly used to extract the time-frequency position features; although our goal is to ignore the time-frequency position information, introducing the position encoder in the initial training stage can help the model understand the characteristics of the time-frequency diagram more comprehensively;

[0077] During the encoder training process, the total loss function L is minimized. total As the training goal, by increasing the weights of the intensity and distribution encoders and reducing the weight of the position encoder, the loss functions of the intensity encoder, distribution encoder and position encoder are reduced, and finally the total loss function converges; at this time, the encoding results E of the intensity encoder, distribution encoder and position encoder are extracted respectively intensity 、E distribution and E shape , and determine a unified representation of radio frequency fingerprint RFF B :

[0078] RFF B =[E intensity ,E distribution ,E position ];

[0079] Among them, E intensity ,E distribution ,E positionRepresent the encoding results of the intensity encoder, distribution encoder and position encoder respectively; during the training process, the weights (α and β) of the intensity and distribution encoders should be rewarded to strengthen the learning of these two encoders. For the position encoder, the penalty weight γ is gradually increased during the training process so that the model tends to ignore the position information. Finally, a unified representation model of RF fingerprints that is resistant to time-frequency resource changes is obtained, so that the encoding results of each encoder are obtained as a unified representation of RF fingerprints.

[0080] The valid RF fingerprint uniformity characterization is confirmed based on the two extracted RF fingerprint features, including:

[0081] A unified representation of RF fingerprint RFF under power histogram is obtained respectively A A unified representation of radio frequency fingerprints under multi-channel autoencoders RFF B , calculate and obtain the agreed radio frequency fingerprint characterization RFF that is resistant to time-frequency resource changes:

[0082] RFF=[RFF A ,RFF B ].

[0083] The sending device has the capability of dynamically allocating time and frequency resources, and sends a baseband signal of a fixed sequence to the signal receiving device.

[0084] An electronic device, comprising:

[0085] at least one processor;

[0086] and, a memory communicatively coupled to the at least one processor;

[0087] The memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that the at least one processor can perform the above-mentioned identification method.

[0088] A computer-readable storage medium stores a computer program, wherein the computer program implements the above-mentioned identification method when executed by a processor.

[0089] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A device identity recognition method, applied to a signal receiving device, characterized in that: The method comprises: Acquire the transmission signals received by the signal receiving device from each sending device under different resource grids to form a transmission signal set; Analyzing the amplitude distribution of the transmission signal at different frequency components according to the transmission signal set; Coarse-grained and fine-grained RF fingerprint extraction are performed respectively to determine a unified RF fingerprint representation that is resistant to time-frequency resource changes.

2. The device identification method according to claim 1, characterized in that: The step of analyzing the amplitude distribution of the transmission signal at different frequency components according to the transmission signal set comprises: Performing preprocessing operations of synchronization, normalization, noise reduction and channel response removal on the transmission signal; Using short-time Fourier transform Convert the preprocessed transmission signal from the time domain to the frequency domain and determine the amplitude variation model of the signal frequency component within the observation time: in, represents the transmission signal under a certain resource grid parameter, ω(t-τ) represents the window function, τ is the time position of the window sliding on the signal, f is the frequency variable, e -j2πft It is the core of Fourier transform, which is used to convert signals from time domain to frequency domain; By changing the width of the window function ω(t-τ), the signal intensity and distribution characteristics of the time-frequency analysis results at different resolutions are used as the representation of the amplitude variation pattern of the signal at different frequency components.

3. The device identification method according to claim 1, characterized in that: The coarse-grained and fine-grained radio frequency fingerprint extractions are respectively performed to determine a unified radio frequency fingerprint representation that is resistant to time-frequency resource changes, including: Obtain the amplitude distribution of the signal frequency components within the observation time, calculate the signal power distribution, and use the frequency of occurrence of different signal powers in each power interval as the RF fingerprint; Design a multi-channel autoencoder structure to extract the intensity, distribution and position features of signal energy in a fine-grained manner through intensity encoder, distribution encoder and position encoder; The valid RF fingerprint uniformity representation is confirmed based on the two extracted RF fingerprint features for device identification.

4. The device identification method according to claim 3, characterized in that: The obtaining of the amplitude distribution of the signal frequency components within the observation time, calculating the signal power distribution and using the frequency of occurrence of different signal powers in each power interval as the radio frequency fingerprint includes: Obtain the amplitude distribution model of the signal frequency component within the observation time Calculate the signal power P: in, and Respectively represent the real part and imaginary part operations; Determine the power histogram H(P) according to the signal power P i ): H(P i )=m i {P|P∈[P i,min ,P i,max ]}; where P i Indicates the set power interval, m i represents the number of power values ​​falling into each power interval, P i,min and P i,max are the minimum and maximum values ​​of the i-th power interval, respectively; The H(P i ) as a unified representation of radio frequency fingerprint RFF A .

5. The device identification method according to claim 3, characterized in that: The multi-channel autoencoder structure is designed to extract the intensity, distribution and position features of signal energy in a fine-grained manner through an intensity encoder, a distribution encoder and a position encoder, including: Obtain the amplitude distribution model of the signal frequency component within the observation time As the input of the multi-channel autoencoder, the total loss function L of the multi-channel autoencoder total Designed for: L total =α(t)·L intensity +β(t)·L distribution +γ(t)·L position ; Among them, L intensity , L distribution and L shape Respectively represent the loss functions of the intensity encoder, distribution encoder and position encoder, α, β and γ represent the weights of the intensity encoder, distribution encoder and position encoder, and their values ​​change dynamically; During the encoder training process, the total loss function L is minimized. total As the training goal, by increasing the weights of the intensity and distribution encoders and reducing the weight of the position encoder, the loss functions of the intensity encoder, distribution encoder and position encoder are reduced, and finally the total loss function converges; at this time, the encoding results E of the intensity encoder, distribution encoder and position encoder are extracted respectively intensity 、E distribution and E shape , and determine a unified representation of radio frequency fingerprint RFF B : RFF B =[E intensity ,It distribution ,It position ]。 6. The device identification method according to claim 3, characterized in that: The valid radio frequency fingerprint uniformity characterization is confirmed based on the two extracted radio frequency fingerprint features, including: A unified representation of RF fingerprint RFF under power histogram is obtained respectively A A unified representation of radio frequency fingerprints under multi-channel autoencoders RFF B , calculate and obtain the agreed radio frequency fingerprint characterization RFF that is resistant to time-frequency resource changes: RFF=[RFF A ,RFF B ]。 7. The device identification method according to claim 1, characterized in that: The sending device has the capability of dynamically allocating time and frequency resources, and sends a baseband signal of a fixed sequence to the signal receiving device.

8. An electronic device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the identification method according to any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the identification method according to any one of claims 1 to 7 is implemented.

Images