Internet of vehicles abnormal equipment identification method based on radio frequency fingerprints and related equipment

Abstract

The invention discloses an Internet of Vehicles abnormal equipment identification method based on radio frequency fingerprints and related equipment, and relates to the technical field of Internet of Vehicles safety. The method comprises the following steps: acquiring a radio frequency signal sent by a to-be-identified device; generating a rotation constellation diagram of the to-be-identified equipment according to a radio frequency signal sent by the to-be-identified equipment; the rotation constellation diagram of the to-be-identified equipment is input into a pre-trained equipment classification model, the equipment classification model comprises a convolutional self-encoding network and a classifier network, and feature extraction is performed on the rotation constellation diagram of the to-be-identified equipment through the convolutional self-encoding network to obtain image feature information; and the image feature information is classified and identified through a classifier network to predict whether the to-be-identified equipment belongs to abnormal equipment which is not allowed to access the Internet of Vehicles. According to the invention, the security of the Internet of Vehicles can be improved.

Description

Technical Field

[0001] This application relates to the field of vehicle network security technology, and in particular to a method and related equipment for identifying abnormal vehicle network devices based on radio frequency fingerprinting. Background Technology

[0002] Existing vehicle network security technologies include encrypted communication, protocol authentication, and intrusion detection. However, these technologies have limitations, resulting in low security for vehicle networks. For example, encrypted communication technology is susceptible to attacks due to poor key management, protocol authentication technology is susceptible to attacks due to protocol design flaws, and intrusion detection technology often relies on detecting preset attack characteristics, making it difficult to cope with constantly changing attack methods. Summary of the Invention

[0003] The main objective of this application is to provide a method and related equipment for identifying abnormal devices in the Internet of Vehicles (IoV) based on radio frequency fingerprinting, aiming to improve the security of IoV.

[0004] To achieve the above objectives, one aspect of this application proposes a method for identifying abnormal devices in the Internet of Vehicles based on radio frequency fingerprinting, comprising the following steps: Acquire the radio frequency signal sent by the device to be identified; A rotating constellation diagram of the device to be identified is generated based on the radio frequency signal sent by the device to be identified. The rotating constellation diagram of the device to be identified is input into a pre-trained device classification model, which includes a convolutional autoencoder network and a classifier network. The convolutional autoencoder network extracts features from the rotating constellation diagram of the device to be identified to obtain image feature information. Then, the classifier network classifies and identifies the image feature information to predict whether the device to be identified is an abnormal device that is not allowed to access the Internet of Vehicles.

[0005] In some embodiments, generating a rotating constellation diagram of the device to be identified based on the radio frequency signal transmitted by the device to be identified includes: All data fields are extracted from the radio frequency signal sent by the device to be identified, and then frequency domain conversion and subcarrier equalization compensation are performed on all data fields to obtain the complex information of all data subcarriers; The complex information of all the data subcarriers is plotted onto the complex plane to generate a rotating constellation diagram of the device to be identified.

[0006] In some embodiments, the radio frequency signal transmitted by the device to be identified includes a plurality of time-domain sampling points; the extraction of all data fields from the radio frequency signal transmitted by the device to be identified includes: Energy detection is performed on the radio frequency signal sent by the device to be identified to remove time-domain sampling points that do not meet the preset conditions, thereby obtaining the first radio frequency signal; Autocorrelation calculation is performed on the first radio frequency signal to determine the positions of all short preambles contained in the first radio frequency signal; Based on the positions of all short preambles contained in the first radio frequency signal and a preset standard long training sequence, the cross-correlation of the first radio frequency signal is calculated to determine the positions of all long preambles contained in the first radio frequency signal. Based on the positions of all long preambles contained in the first radio frequency signal, all data fields are extracted from the first radio frequency signal.

[0007] In some embodiments, performing frequency domain transformation and subcarrier equalization compensation on all data fields to obtain complex information of all data subcarriers includes: Perform a Fast Fourier Transform on all the data fields to obtain all the first data fields represented in the frequency domain; For each data field and its corresponding first data field, the longest preamble closest to the data field is obtained from the first radio frequency signal and recorded as the target long preamble. Gain is calculated based on the target long preamble and a preset standard polarity sequence to obtain the channel gain sequence corresponding to the data field. Then, based on a preset subcarrier mapping table and the channel gain sequence corresponding to the data field, subcarrier equalization compensation is performed on the first data field, and the complex information of all data subcarriers is extracted.

[0008] In some embodiments, plotting the complex information of all data subcarriers onto the complex plane to generate a rotating constellation diagram of the device to be identified includes: The complex information of all data subcarriers is stored in a preset array, and then the complex information of all data subcarriers is drawn onto the complex plane according to the storage position of the complex information of all data subcarriers in the preset array to generate a rotating constellation diagram of the device to be identified.

[0009] In some embodiments, the device classification model is trained in the following manner: Obtain a first dataset carrying normal category labels and a second dataset carrying abnormal category labels; wherein, the first dataset contains a rotating constellation diagram of multiple normal devices, the normal devices being devices allowed to access the vehicle network, and the rotating constellation diagram of the normal devices is generated based on the radio frequency signals transmitted by the normal devices; the second dataset contains a rotating constellation diagram of multiple abnormal devices, and the rotating constellation diagram of the abnormal devices is generated based on the radio frequency signals transmitted by the abnormal devices. The first dataset and the second dataset are input into a pre-built device classification model for iterative training. In each training round, the classification loss is calculated based on the first dataset and the second dataset, and the model parameters are optimized based on the classification loss. At the end of the iterative training, the trained device classification model is obtained.

[0010] In some embodiments, calculating the classification loss based on the first dataset and the second dataset includes: The convolutional autoencoder network included in the pre-built device classification model extracts features from the first dataset and the second dataset respectively, resulting in a first image feature set and a second image feature set; then, the classifier network included in the pre-built device classification model classifies and identifies the first image feature set and the second image feature set respectively, resulting in a first classification result set and a second classification result set. The first loss is obtained by calculating cross-entropy loss based on the first classification result set, the second classification result set, the normal category label carried by the first dataset, and the abnormal category label carried by the second dataset; and the second loss is obtained by calculating triplet loss based on the first image feature set and the second image feature set. The first loss and the second loss are weighted and summed to obtain the classification loss.

[0011] To achieve the above objectives, another aspect of this application proposes a vehicle-to-everything (V2X) device identification device based on radio frequency fingerprinting, the device comprising: The acquisition module is used to acquire the radio frequency signal sent by the device to be identified; The generation module is used to generate a rotating constellation diagram of the device to be identified based on the radio frequency signal sent by the device to be identified; The classification and recognition module is used to input the rotated constellation diagram of the device to be identified into a pre-trained device classification model. The device classification model includes a convolutional autoencoder network and a classifier network. The convolutional autoencoder network extracts features from the rotated constellation diagram of the device to be identified to obtain image feature information. Then, the classifier network classifies and recognizes the image feature information to predict whether the device to be identified belongs to an abnormal device that is not allowed to access the Internet of Vehicles.

[0012] To achieve the above objectives, another aspect of this application proposes an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method for identifying abnormal devices in the Internet of Vehicles based on radio frequency fingerprinting.

[0013] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for identifying abnormal devices in the Internet of Vehicles based on radio frequency fingerprints.

[0014] The embodiments of this application include at least the following beneficial effects: by converting the radio frequency signal sent by the device to be identified into a rotating constellation diagram that can more intuitively express the inherent characteristics of the device hardware and inputting it into the device classification model for classification and identification, the impact of signal noise and channel interference on the classification and identification results can be reduced, thereby more reliably predicting whether the device to be identified belongs to an abnormal device that is not allowed to access the vehicle network, thereby improving the security of the vehicle network.

[0015] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0016] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart illustrating a method for identifying abnormal devices in a vehicle network based on radio frequency fingerprinting, as provided in an embodiment of this application. Figure 2 This is a schematic diagram of the rotating constellation diagram provided in the embodiments of this application; Figure 3 This is a schematic diagram of the composition of a vehicle networking abnormal device identification device based on radio frequency fingerprinting provided in an embodiment of this application; Figure 4 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the reference to "embodiment" herein means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. The implementation methods described in the following exemplary embodiments do not represent all implementation methods consistent with the embodiments of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0018] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0019] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0021] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0022] With the rapid development of vehicle-to-everything (V2X) and wireless communication technologies, the application of wireless technology in V2X is becoming increasingly widespread. Vehicles can integrate functions such as navigation, remote diagnostics, emergency assistance, and entertainment information services by wirelessly communicating with external devices and services, greatly enhancing the user's driving experience and the vehicle's intelligence level. However, the openness of this wireless connection also brings certain security risks. Illegal devices may use methods such as spoofing or cracking wireless communication protocols to infiltrate the V2X system, thereby controlling the vehicle or stealing sensitive information, causing property damage to the vehicle owner or even threatening personal safety.

[0023] Existing vehicle network security technologies include encrypted communication, protocol authentication, and intrusion detection. However, these technologies have limitations, resulting in low security for vehicle networks. For example, encrypted communication technology is susceptible to attacks due to poor key management, protocol authentication technology is susceptible to attacks due to protocol design flaws, and intrusion detection technology often relies on detecting preset attack characteristics, making it difficult to cope with constantly changing attack methods.

[0024] Radio frequency fingerprinting (RF fingerprinting) technology, as an emerging wireless device identification technology, can distinguish different wireless devices by utilizing the unique radio frequency signal characteristics generated by the inherent characteristics of the device hardware. This technology has the characteristics of not requiring modification of device hardware and being difficult to forge and copy, and has great potential in the field of wireless network security. How to effectively apply RF fingerprinting technology to the field of abnormal device identification in vehicle networking is a technical challenge that engineers have been committed to researching.

[0025] In view of this, this application proposes a method and related equipment for identifying abnormal devices in the Internet of Vehicles based on radio frequency fingerprinting. This scheme converts the radio frequency signal sent by the device to be identified into a rotating constellation diagram that can more intuitively express the inherent characteristics of the device hardware and inputs it into the device classification model for classification and identification. This can reduce the impact of signal noise and channel interference on the classification and identification results, and thus more reliably predict whether the device to be identified belongs to an abnormal device that is not allowed to access the Internet of Vehicles, thereby improving the security of the Internet of Vehicles.

[0026] This application provides a method for identifying abnormal devices in the Internet of Vehicles (IoV) based on radio frequency fingerprinting. This method can be applied to the electronic devices provided in this application, which can be terminals or servers. Terminals can be tablets, laptops, desktop computers, etc., but are not limited to these. Servers can be independent physical servers, server clusters or distributed systems composed of multiple physical servers, or cloud servers providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and big data and artificial intelligence platforms.

[0027] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for identifying abnormal devices in a vehicle network based on radio frequency fingerprinting, as provided in an embodiment of this application. It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0028] The method for identifying abnormal devices in the Internet of Vehicles based on radio frequency fingerprinting provided in this application may include, but is not limited to, the three steps S101 to S103, as detailed below: S101. Acquire the radio frequency signal sent by the device to be identified; S102. Generate a rotating constellation diagram of the device to be identified based on the radio frequency signal sent by the device to be identified; S103. Input the rotated constellation diagram of the device to be identified into a pre-trained device classification model. The device classification model includes a convolutional autoencoder network and a classifier network. The convolutional autoencoder network extracts features from the rotated constellation diagram of the device to be identified to obtain image feature information. Then, the classifier network classifies and identifies the image feature information to predict whether the device to be identified belongs to an abnormal device that is not allowed to access the Internet of Vehicles.

[0029] The three steps S101 to S103 shown in the embodiments of this application convert the radio frequency signal sent by the device to be identified into a rotating constellation diagram and input it into the device classification model for classification and identification. This can more reliably predict whether the device to be identified belongs to an abnormal device that is not allowed to access the vehicle network, thereby improving the security of the vehicle network.

[0030] In some embodiments, the method for generating the rotating constellation diagram of the device to be identified in S102 may include, but is not limited to, the two steps S201 to S202, as follows: S201. Extract all data fields from the radio frequency signal sent by the device to be identified, and then perform frequency domain conversion and subcarrier equalization compensation on all data fields to obtain the complex information of all data subcarriers. S202. Plot the complex information of all data subcarriers onto the complex plane to generate a rotating constellation diagram of the device to be identified.

[0031] In S201 above, the radio frequency signal transmitted by the device to be identified contains several time-domain sampling points, each of which can be represented as: , , ; In the formula, Refers to time-domain sampling points, which are represented in complex form. For the in-phase components of the time-domain sampling points, These are the orthogonal components of the time-domain sampling points. The imaginary unit, For amplitude, Angular frequency, The sampling interval is... This is the initial phase.

[0032] Based on this, the corresponding implementation of the step of extracting all data fields from the radio frequency signal sent by the device to be identified may include, but is not limited to, the following four steps S301 to S304, as detailed below: S301. Perform energy detection on the radio frequency signal sent by the device to be identified to remove time-domain sampling points that do not meet the preset conditions, thereby forming a first radio frequency signal; wherein, the preset conditions are used to limit the in-phase component of the time-domain sampling points to be greater than a preset threshold, which can be adopted as follows: This expression is used to represent it. For the in-phase components of the time-domain sampling points, The preset threshold; S302. Perform autocorrelation calculation on the first radio frequency signal to determine the positions of all short preambles contained in the first radio frequency signal. Each short preamble consists of 10 identical short training sequences, and each short training sequence contains 16 time-domain sampling points represented in complex form. The autocorrelation calculation formula is as follows: ; In the formula, For time domain sampling points Similarity values ​​between itself and its own values ​​at different offsets The window size is set to 16 in this application. for The conjugate representation of , Refers to the sampling point in the time domain The 16th time-domain sampling point, for The conjugate representation of the similarity value obtained by calculation When the similarity is greater than or equal to the preset similarity threshold, it indicates that the time-domain sampling points Belongs to the first shortest training sequence, time-domain sampling points It belongs to the second shortest training sequence, which immediately follows the first shortest training sequence. S303. Based on the positions of all short preambles contained in the first radio frequency signal and a preset standard long training sequence, calculate the cross-correlation of the first radio frequency signal to determine the positions of all long preambles contained in the first radio frequency signal. Each long preamble follows a short preamble, each long preamble contains two identical long training sequences, and each long training sequence contains 64 time-domain sampling points represented in complex form. The cross-correlation calculation formula is as follows: ; In the formula, For time-domain sampling points The cross-correlation value between the sequence consisting of the following 63 time-domain sampling points and the standard long preamble, when the calculated cross-correlation value... When the cross-correlation is greater than or equal to the preset cross-correlation threshold, it indicates that the cross-correlation is generated by the time-domain sampling points. The sequence consisting of the first radio frequency signal and the following 63 time-domain sampling points constitutes a long preamble. Refers to the sampling point in the time domain After that Bit time-domain sampling points, This refers to the standard long preamble containing elements falling within the first... The standard time-domain sampling points are also represented in complex form. for The conjugate representation; S304. Based on the positions of all long preambles contained in the first radio frequency signal, extract all data fields from the first radio frequency signal. Each data field falls after a long preamble, and each data field contains 64 time-domain sampling points represented in complex form.

[0033] In the above S201, the step of performing frequency domain transformation and subcarrier equalization compensation on all data fields to obtain the complex information of all data subcarriers can be implemented in a way that includes, but is not limited to, the following two steps S401 to S402, as follows: S401. Perform a Fast Fourier Transform on all data fields represented in the time domain to obtain all first data fields represented in the frequency domain; wherein, the Fast Fourier Transform formula is: ; In the formula, for each data field and the first data field obtained after performing a fast Fourier transform on it, For this data field, the data that falls within the first... Bit time-domain sampling points, For the first data field to contain the data falling within the first... Frequency domain sampling points of bits For sampling time index, Number the frequency components; S402. For each data field and the first data field obtained after performing a Fast Fourier Transform, first obtain the longest preamble closest to the data field from the first radio frequency signal and record it as the target long preamble. The target long preamble falls before the data field. Calculate the gain based on the target long preamble and a preset standard polarity sequence to obtain the channel gain sequence corresponding to the data field; wherein, the gain calculation formula is: ; ; In the formula, and The target long preamble contains two long training sequences. This refers to the sequence numbered after the long training sequence at the beginning has been transformed from the time domain to the frequency domain representation. Frequency domain sampling points, Refers to the sequence of training data that follows after the time-domain to frequency-domain representation is transformed, and includes the numbered sequences. Frequency domain sampling points, The standard polarity sequence contains the numbered The polarity value of the standard polarity sequence can characterize the polarity of the standard-length training sequence. The channel gain sequence contains the numbered The channel gain, these two long training sequences, the standard polarity sequence, the channel gain sequence and the first data field all assign the data in the middle position to 0, and starting from the data in the middle position, assign decreasing negative integer numbers to the data to the left of the middle position, and assign increasing positive integer numbers to the data to the right of the middle position. Then, based on the preset subcarrier mapping table and the channel gain sequence corresponding to the data field, the complex information of all data subcarriers is extracted after subcarrier equalization compensation is performed on the first data field represented in the frequency domain; wherein, the subcarrier equalization compensation formula is: ; In the formula, This refers to the number contained in the first data field. The frequency domain sampling points can be understood as the subcarrier equalization compensation formula only applying subcarrier equalization compensation to the frequency domain sampling points numbered -26 to 26 contained in the first data field. This refers to the number contained in the first data field. frequency domain sampling points The corresponding subcarrier equalization compensation result, and when the first data field contains the number as frequency domain sampling points When the corresponding data subcarrier is used, the equalization compensation result of the corresponding subcarrier is used. Complex information is stored as data subcarriers.

[0034] The subcarrier mapping table reflects the subcarrier structure specified according to the 802.11 protocol standard, which can be found in [reference needed]. Figure 2As shown, this subcarrier structure contains 64 subcarriers. The subcarrier in the middle position is defined as a DC subcarrier, numbered 0, and is usually not used for data transmission to avoid DC offset interference. The subcarriers 7 positions before, 21 positions before, 7 positions after, and 21 positions after the DC subcarrier are defined as pilot subcarriers, numbered -7, -21, 7, and 21 respectively, and are used for channel estimation and carrier synchronization. The 6 subcarriers 1 to 6 positions before the DC subcarrier, the 13 subcarriers 8 to 20 positions before the DC subcarrier, and the subcarriers 22 to 21 positions before the DC subcarrier are all defined as pilot subcarriers, numbered -7, -21, 7, and 21 respectively, and are used for channel estimation and carrier synchronization. The five subcarriers at position 26, the six subcarriers at positions 1 to 6 following the DC subcarrier, the thirteen subcarriers at positions 8 to 20 following the DC subcarrier, and the five subcarriers at positions 22 to 26 following the DC subcarrier are all defined as data subcarriers, with corresponding numbers -1 to -6, -8 to -20, -22 to -26, 1 to 6, 8 to 20, and 22 to 26, and are typically used for data transmission. The five subcarriers at positions 27 to 31 before the DC subcarrier and the six subcarriers at positions 27 to 32 following the DC subcarrier are all defined as guard band subcarriers, with corresponding numbers -27 to -32 and 27 to 31, and are used to reduce out-of-band radiation and adjacent channel interference.

[0035] In the above S202, the complex information of all data subcarriers can be stored in a preset array first, and then the complex information of all data subcarriers can be drawn on the complex plane according to the storage position of the complex information of all data subcarriers in the preset array to generate a rotating constellation diagram of the device to be identified.

[0036] Specifically, the preset array contains several storage locations, numbered sequentially from 0. Each storage location can only store the complex information of one data subcarrier. The complex plane has a real axis and an imaginary axis; the real axis corresponds to the in-phase component, and the imaginary axis corresponds to the quadrature component. For the complex information of each data subcarrier, it includes the in-phase and quadrature components carried by that data subcarrier. The number corresponding to the storage location of the complex information of that data subcarrier in the preset array is obtained. The in-phase and quadrature components carried by the data subcarrier are multiplied by the square of this number, and then combined to form a signal point coordinate, which is then plotted on the complex plane. When all signal point coordinates obtained from the transformation based on the complex information of all data subcarriers are plotted on the complex plane, a rotating constellation diagram of the device to be identified can be obtained. (See [reference]). Figure 2 As shown.

[0037] In this application, by converting the radio frequency signals transmitted by the device into a rotating constellation diagram as model input, the advantages of computer vision technology can be fully utilized for image feature extraction and classification, which is beneficial to improving model learning efficiency and reducing model computational costs. Furthermore, the rotating constellation diagram of the device can effectively reveal hardware differences and reduce the adverse effects of signal noise and channel interference on model classification and recognition.

[0038] In some embodiments, the training method for the device classification model in S103 may include, but is not limited to, steps S501 to S502, as follows: S501. Obtain a first dataset carrying a normal category label and a second dataset carrying an abnormal category label; wherein, the first dataset contains rotating constellation diagrams of multiple normal devices, which are devices allowed to access the vehicle network, and the rotating constellation diagrams of normal devices are generated based on the radio frequency signals transmitted by the normal devices; the second dataset contains rotating constellation diagrams of multiple abnormal devices, which are devices not allowed to access the vehicle network, and the rotating constellation diagrams of abnormal devices are generated based on the radio frequency signals transmitted by the abnormal devices. S502. Input the first and second datasets into the pre-built device classification model for iterative training. In each training round, first calculate the classification loss based on the first and second datasets, then optimize the model parameters based on the classification loss. At the end of the iterative training, the trained device classification model is obtained. The model parameter optimization process can be implemented using the gradient descent method. The termination condition for model iterative training can be reaching the preset maximum number of iterations or the classification loss reaching a convergent state.

[0039] In step S501 above, the first dataset may further include multiple first data subsets corresponding to multiple normal devices. Each first data subset corresponding to a normal device contains multiple rotating constellation diagrams generated based on multiple radio frequency signals transmitted by that normal device within several different time periods. Similarly, the second dataset may further include multiple second data subsets corresponding to multiple abnormal devices. Each second data subset corresponding to an abnormal device contains multiple rotating constellation diagrams generated based on multiple radio frequency signals transmitted by that abnormal device within several different time periods. Preferably, the number of normal devices and the number of abnormal devices are both set to 8. It should be noted that any device can adjust its receiving distance at different time periods, which facilitates the generation of other more distinctive rotating constellation diagrams.

[0040] It should be noted that the generation method for the rotating constellation diagram of each normal device and each abnormal device is the same as the generation method for the rotating constellation diagram of the device to be identified, and will not be repeated here.

[0041] It should be noted that devices allowed to access the vehicle network can be understood as authorized or legally connected devices, while devices not allowed to access the vehicle network can be understood as unauthorized or illegally connected devices.

[0042] In the above S502, the pre-built device classification model includes a convolutional autoencoder network and a classifier network. An encoder-decoder architecture is introduced in the convolutional autoencoder network, and a sigmoid layer is introduced in the classifier network. The convolutional autoencoder network is used to extract features from the rotating constellation diagram of the input device. That is, the rotating constellation diagram is gradually transformed into a denoised low-dimensional feature vector through multi-layer convolution and pooling operations. Then, the classifier network is used to classify and identify the extracted image feature information. The obtained prediction classification result is used to reflect whether the device belongs to a normal device that is allowed to access the vehicle network or an abnormal device that is not allowed to access the vehicle network.

[0043] In this application, signal noise can be effectively removed through an autoencoder process, and representative feature vectors can be extracted from the rotating constellation diagram of the input device. In particular, in the processing of rotating constellation diagrams of devices with small differences in hardware characteristics, more representative feature information can be learned through the reconstruction of the decoder.

[0044] In S502 above, the step of calculating the classification loss based on the first dataset and the second dataset can be implemented in ways that include, but are not limited to, the following three steps: S601 to S603, as detailed below: S601. The first dataset and the second dataset are extracted using the convolutional autoencoder network contained in the pre-built device classification model, respectively, to obtain the first image feature set and the second image feature set; then the first image feature set and the second image feature set are classified and recognized using the classifier network contained in the pre-built device classification model, respectively, to obtain the first classification result set and the second classification result set. S602. Calculate the first loss by performing cross-entropy loss based on the first classification result set corresponding to the first dataset, the normal category label carried by the first dataset, the second classification result set corresponding to the second dataset, and the abnormal category label carried by the second dataset; wherein, the cross-entropy loss calculation formula is: ; In the formula, The first loss, The number of all rotated constellation diagrams contained in the first and second datasets. For the first The true category labels carried by a rotating constellation chart. For the first The predicted probability of a rotating constellation being predicted as a positive class, where the positive class refers to the true class of the rotating constellation; And the second loss is obtained by calculating the triplet loss based on the first image feature set corresponding to the first dataset and the second image feature set corresponding to the second dataset; S603. The first loss and the second loss are weighted and summed to obtain the classification loss.

[0045] In step S601 above, for each normal device's rotating constellation diagram in the first dataset, the first image feature information is obtained by extracting features from the rotating constellation diagram of the normal device using the convolutional autoencoder network included in the pre-built device classification model. Then, the first image feature information is classified and identified using the classifier network included in the pre-built device classification model to obtain a first classification result. The first classification result includes each predicted category of the normal device and the prediction probability corresponding to each predicted category. Similarly, for each abnormal device's rotating constellation diagram in the second dataset, the second image feature information is obtained by extracting features from the rotating constellation diagram of the abnormal device using the convolutional autoencoder network included in the pre-built device classification model. Then, the second image feature information is classified and identified using the classifier network included in the pre-built device classification model to obtain a second classification result. The second classification result includes each predicted category of the abnormal device and the prediction probability corresponding to each predicted category.

[0046] In S602 above, the step of calculating the triplet loss based on the first image feature set corresponding to the first dataset and the second image feature set corresponding to the second dataset to obtain the second loss can be implemented in ways that include, but are not limited to, the following: Samples are extracted from the first and second datasets to form multiple triplet samples. Each triplet sample contains an anchor sample, a positive sample, and a negative sample. The triplet sample can primarily come from the first dataset; that is, the anchor sample and the positive sample come from the rotation constellation diagrams of any two normal devices included in the first dataset, and the negative sample comes from the rotation constellation diagram of any one abnormal device included in the second dataset. Alternatively, the triplet sample can primarily come from the second dataset; that is, the anchor sample and the positive sample come from the rotation constellation diagrams of any two abnormal devices included in the second dataset, and the negative sample comes from the rotation constellation diagram of any one normal device included in the first dataset. Triplet loss is calculated based on all image feature information corresponding to each triplet sample to obtain the triplet loss value. The average of the multiple triplet loss values ​​corresponding to multiple triplet samples is then calculated to obtain the second loss. The formula for calculating the triplet loss is as follows: ; In the formula, The triplet loss value, Refers to anchor point samples, Refers to positive samples. Refers to negative samples, This is a mapping function used to map samples to a feature space. The image feature information corresponding to the anchor point sample. The image feature information corresponding to the positive samples. Image feature information for negative samples. The boundary is used to define positive samples. and negative samples The minimum distance that should exist between them in the feature space can be understood as a hyperparameter. It refers to the L2 norm, which is the square of the Euclidean distance.

[0047] In this application, by supervising the training of a pre-built device classification model based on a first dataset and a second dataset, the device classification model can be made to have reliable classification and recognition performance.

[0048] In this application, in order to comprehensively evaluate the performance of the trained device classification model in real-world application tasks, a series of performance metrics can be used for testing, including accuracy, precision, and recall.

[0049] Please refer to Figure 3 , Figure 3 This is a schematic diagram of a vehicle-to-everything (V2X) device identification device based on radio frequency fingerprinting, provided in an embodiment of this application. It can implement the aforementioned V2X device identification method based on radio frequency fingerprinting. The device may include, but is not limited to, the following: The acquisition module 701 is used to acquire the radio frequency signal sent by the device to be identified; The generation module 702 is used to generate a rotating constellation diagram of the device to be identified based on the radio frequency signal sent by the device to be identified. The classification and recognition module 703 is used to input the rotated constellation diagram of the device to be identified into a pre-trained device classification model. The device classification model includes a convolutional autoencoder network and a classifier network. The convolutional autoencoder network extracts features from the rotated constellation diagram of the device to be identified to obtain image feature information. Then, the classifier network classifies and recognizes the image feature information to predict whether the device to be identified belongs to an abnormal device that is not allowed to access the Internet of Vehicles.

[0050] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those specifically implemented by the above method embodiments, and the beneficial effects achieved by the present device embodiments are also the same as those achieved by the above method embodiments.

[0051] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method for identifying abnormal devices in a vehicle network based on radio frequency fingerprinting. This electronic device can include any smart terminal such as a tablet computer or an in-vehicle computer.

[0052] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those implemented by the above method embodiments, and the beneficial effects achieved by the present device embodiments are also the same as those achieved by the above method embodiments.

[0053] Please see Figure 4 , Figure 4 This is a schematic diagram illustrating the hardware structure of an electronic device according to another embodiment. The electronic device includes: The processor 801 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 802 can be implemented in the form of read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 802 can store the operating system and other applications. When the technical solution provided in the embodiments of this application is implemented by software or firmware, the relevant program code is stored in the memory 802 and is called and executed by the processor 801. The 803 input / output interface is used to implement information input and output. The communication interface 804 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 805 transmits information between various components of the device (e.g., processor 801, memory 802, input / output interface 803, and communication interface 804); The processor 801, memory 802, input / output interface 803, and communication interface 804 are connected to each other within the device via bus 805.

[0054] This application also provides a computer program product, which includes a computer program that, when executed by one or more processors, implements the above-described method for identifying abnormal devices in the Internet of Vehicles based on radio frequency fingerprinting.

[0055] It is understood that the content of the above method embodiments is applicable to this computer program product. The specific functions implemented by the embodiments of this computer program product are the same as those implemented by the above method embodiments, and the beneficial effects achieved by the embodiments of this computer program product are also the same as those achieved by the above method embodiments.

[0056] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0057] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0058] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0059] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0060] Those skilled in the art will understand that all or some of the steps, apparatuses, or functional modules / units in the methods disclosed above can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0061] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0062] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0063] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed between the devices or units may be through some interfaces, and the indirect coupling or communication connection may be electrical, mechanical, or other forms.

[0064] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0065] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0067] The preferred embodiments of this application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of this application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of this application shall be within the scope of the claims of this application.

Claims

1. A radio frequency fingerprint based identification of abnormal devices in Internet of Vehicles, characterized in that, The method comprises the following steps: Obtaining a radio frequency signal sent by a to-be-identified device; Generating a rotating constellation diagram of the to-be-identified device according to the radio frequency signal sent by the to-be-identified device; Inputting the rotating constellation diagram of the to-be-identified device into a pre-trained device classification model, wherein the device classification model comprises a convolutional auto-encoding network and a classifier network, the convolutional auto-encoding network is used to extract features of the rotating constellation diagram of the to-be-identified device to obtain image feature information, and the classifier network is used to classify and identify the image feature information to predict whether the to-be-identified device belongs to an abnormal device that is not allowed to access the Internet of Vehicles. 2.The radio frequency fingerprint based IoT anomaly device identification method of claim 1, wherein, The generating of the rotating constellation diagram of the to-be-identified device according to the radio frequency signal sent by the to-be-identified device comprises: Extracting all data fields from the radio frequency signal sent by the to-be-identified device, and performing frequency domain conversion and subcarrier equalization compensation on the all data fields to obtain complex information of all data subcarriers; Drawing the complex information of all data subcarriers to a complex plane to generate the rotating constellation diagram of the to-be-identified device. 3.The radio frequency fingerprint based abnormal device identification method for Internet of Vehicles according to claim 2, characterized in that, The radio frequency signal sent by the to-be-identified device comprises a plurality of time domain sampling points; the extracting of all data fields from the radio frequency signal sent by the to-be-identified device comprises: Performing energy detection on the radio frequency signal sent by the to-be-identified device to eliminate time domain sampling points that do not meet preset conditions to obtain a first radio frequency signal; Performing autocorrelation calculation on the first radio frequency signal to determine positions of all short preambles contained in the first radio frequency signal; Performing cross-correlation calculation on the first radio frequency signal according to the positions of all short preambles contained in the first radio frequency signal and a preset standard long training sequence to determine positions of all long preambles contained in the first radio frequency signal; Extracting the all data fields from the first radio frequency signal according to the positions of all long preambles contained in the first radio frequency signal. 4.The radio frequency fingerprint based abnormal device identification method for Internet of Vehicles according to claim 3, characterized in that, The frequency domain conversion and subcarrier equalization compensation on the all data fields to obtain complex information of all data subcarriers comprise: Performing fast Fourier transform on the all data fields to correspondingly obtain all first data fields represented in the frequency domain; For each data field and the corresponding first data field, obtaining a long preamble closest to the data field from the first radio frequency signal and recording the long preamble as a target long preamble, performing gain calculation according to the target long preamble and a preset standard polarity sequence to obtain a channel gain sequence corresponding to the data field, and extracting the complex information of all data subcarriers after performing subcarrier equalization compensation on the first data field according to a preset subcarrier mapping table and the channel gain sequence corresponding to the data field. 5.The radio frequency fingerprint based IoT anomaly device identification method of claim 2, wherein, The drawing of the complex information of all data subcarriers to a complex plane to generate the rotating constellation diagram of the to-be-identified device comprises: The complex information of all the data subcarriers is stored in a preset array, and the complex information of all the data subcarriers is plotted on the complex plane according to the storage positions of the complex information of all the data subcarriers in the preset array, so as to generate a rotating constellation diagram of the device to be identified. 6.The radio frequency fingerprint-based abnormal device identification method for Internet of Vehicles according to claim 1, wherein, The device classification model is trained in the following manner: A first data set carrying normal category labels and a second data set carrying abnormal category labels are obtained; the first data set contains rotating constellation diagrams of multiple normal devices, the normal devices being devices allowed to access the vehicle network, the rotating constellation diagrams of the normal devices being generated based on radio frequency signals sent by the normal devices, and the second data set containing rotating constellation diagrams of multiple abnormal devices, the rotating constellation diagrams of the abnormal devices being generated based on radio frequency signals sent by the abnormal devices; The first data set and the second data set are input into a pre-constructed device classification model for iterative training, and in each round of training, a classification loss is calculated according to the first data set and the second data set, and model parameter optimization is performed according to the classification loss, so as to obtain a trained device classification model at the end of the iterative training. 7.The radio frequency fingerprint-based abnormal device identification method for Internet of Vehicles according to claim 6, characterized in that, The classification loss is calculated according to the first data set and the second data set, which includes: The first data set and the second data set are respectively subjected to feature extraction by a convolutional auto-encoder network included in the pre-constructed device classification model, so as to obtain a first image feature set and a second image feature set; and the first image feature set and the second image feature set are respectively subjected to classification recognition by a classifier network included in the pre-constructed device classification model, so as to obtain a first classification result set and a second classification result set; A first loss is obtained by cross-entropy loss calculation according to the first classification result set, the second classification result set, normal category labels carried by the first data set, and abnormal category labels carried by the second data set; and a second loss is obtained by triple loss calculation according to the first image feature set and the second image feature set; The first loss and the second loss are weighted and summed to obtain the classification loss. 8.A radio frequency fingerprint based device identification apparatus for Internet of Vehicles, characterized in that, The device includes: An acquisition module configured to acquire a radio frequency signal sent by a device to be identified; A generation module configured to generate a rotating constellation diagram of the device to be identified according to the radio frequency signal sent by the device to be identified; A classification and recognition module configured to input the rotating constellation diagram of the device to be identified into a pre-trained device classification model, the device classification model including a convolutional auto-encoder network and a classifier network, the convolutional auto-encoder network being used to extract feature information of the rotating constellation diagram of the device to be identified, and the classifier network being used to classify and recognize the feature information to predict whether the device to be identified belongs to an abnormal device not allowed to access a vehicle network.

9. An electronic device, comprising: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the radio frequency fingerprint-based vehicle networking abnormal device identification method in any one of claims 1 to 7 when executing the computer program.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the radio frequency fingerprint-based vehicle networking abnormal device identification method in any one of claims 1 to 7.

Images