Federal learning radio frequency fingerprint identification method and system based on attention feature fusion

Abstract

The invention provides a federal learning radio frequency fingerprint identification method and system based on attention feature fusion. The method comprises the following steps: a plurality of clients respectively obtain corresponding radio frequency signals as training data sets; each client carries out training on the lightweight radio frequency fingerprint identification model according to the corresponding training data set so as to update local parameters, and sends the local parameters corresponding to each client to the server; the server side carries out weighted aggregation on the received local parameters and sends the local parameters as global parameters to each client side to carry out iterative training, and in the iterative training, each client side adopts an attention feature fusion module to fuse the global parameters received in the current round and the local parameters updated in the last round of training so as to carry out personalized adjustment; the server side adopts an attention feature fusion module to fuse the weighted and aggregated parameters and the global parameters of the previous round to obtain the global parameters of the current round; therefore, the radio frequency fingerprint identification accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of wireless communication security technology, and in particular to a federated learning radio frequency fingerprinting method based on attention feature fusion, a federated learning radio frequency fingerprinting system based on attention feature fusion, a computer-readable storage medium, and a computer device. Background Technology

[0002] With the continuous development of wireless communication technology, the application scenarios of wireless communication devices have expanded from smartphones and smart homes to fields such as the Industrial Internet of Things (IIoT), the Internet of Vehicles (IoV), and smart manufacturing. However, with the continuous opening of wireless communication network architecture and the exponential growth in the scale of connected wireless communication devices, the information security threats such as identity forgery, spoofing attacks, and replay attacks are constantly increasing. Therefore, how to effectively achieve identity authentication of wireless devices and prevent malicious attacks has become a key issue that needs to be addressed in the field of wireless communication security.

[0003] Traditional radio frequency (RF) fingerprinting technology, based on manual feature selection, primarily extracts features by analyzing the waveform, spectrum, and phase characteristics of RF signals. While this manual feature extraction method can capture some key RF fingerprint information, it is cumbersome, time-consuming, and reliant on prior knowledge. Furthermore, the limitations of manually extracted features typically affect the generalization ability of RF fingerprinting. Deep learning-based RF fingerprinting technology has become a research hotspot in recent years. It extracts deeper and more discriminative fingerprint information from the original RF signal by constructing neural networks with multiple hidden layers.

[0004] Compared to traditional RFID fingerprinting technology, deep learning-based technology offers higher recognition accuracy. However, its practical application in edge embedded scenarios presents two main challenges. First, existing deep learning-based RFID fingerprinting technologies typically require substantial storage and computing resources, placing a significant burden on resource-constrained edge embedded devices and hindering the effective deployment of deep learning network models. Second, existing centralized RFID fingerprinting technologies require edge embedded devices to upload all their local RFID signal data to a central server for centralized data processing and model training. This process consumes significant communication resources, places a heavy burden on the central server's storage space, and poses a risk of data privacy breaches. To address this, federated learning has been introduced to achieve distributed RFID fingerprinting. However, heterogeneity issues exist among the local RFID signal datasets of various edge embedded devices, such as inconsistent signal categories and sample size distributions, as well as differences in receivers. This leads to poor fingerprint recognition performance of the federated trained model. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in the aforementioned technologies. To this end, one objective of this invention is to propose a federated learning radio frequency fingerprinting method based on attention feature fusion, which can effectively mitigate the impact of data heterogeneity on model performance. The model after convergence of federated training exhibits high radio frequency fingerprinting accuracy and is suitable for deployment in edge distributed scenarios.

[0006] A second objective of this invention is to provide a computer-readable storage medium.

[0007] The third objective of this invention is to provide a computer device.

[0008] The fourth objective of this invention is to propose a federated learning radio frequency fingerprint recognition system based on attention feature fusion.

[0009] To achieve the above objectives, a first aspect of this invention proposes a federated learning radio frequency fingerprinting method based on attention feature fusion. This method includes the following steps: multiple clients acquire corresponding radio frequency signals as training datasets; each client trains a lightweight radio frequency fingerprinting model based on depthwise separable grouped convolutions using its corresponding training dataset to update local parameters and sends the local parameters to the server; the server performs weighted aggregation of the received local parameters based on the training sample size of each client and sends the aggregated parameters as global parameters to each client for iterative training. In each round of iterative training, each client uses an attention feature fusion module to fuse the global parameters received in the current round with the updated local parameters from the previous round for personalized adjustments; the server uses the attention feature fusion module to fuse the weighted aggregated parameters with the global parameters from the previous round to obtain the global parameters for the current round. This effectively mitigates the impact of data heterogeneity on model performance, and the model after federated training converges has high radio frequency fingerprinting accuracy, making it suitable for deployment in edge-distributed scenarios.

[0010] In addition, the federated learning RF fingerprinting method based on attention feature fusion proposed in the above embodiments of the present invention may also have the following additional technical features:

[0011] Optionally, the lightweight RF fingerprint recognition model includes an RNN denoising module, a one-dimensional convolutional layer, grouped convolutions, depthwise separable convolutions, a one-dimensional max pooling layer, and a fully connected layer.

[0012] Optionally, the lightweight RF fingerprint recognition model uses an RNN denoising module to denoise the input RF signal to obtain a denoised signal. The denoised signal is then processed by a one-dimensional convolutional layer, grouped convolution, and depthwise separable convolution to extract features, gradually reducing the spatial dimension of the features. Finally, the extracted features are mapped to the category dimension through a one-dimensional max pooling layer and a fully connected layer to construct the relationship between features and categories.

[0013] Optionally, the lightweight RF fingerprint recognition model uses the Adam optimizer and cross-entropy loss function to optimize parameters during training, with a learning rate of 0.001, momentum of 0.9, and 100 training iterations.

[0014] To achieve the above objectives, a second aspect of the present invention provides a computer-readable storage medium storing a federated learning radio frequency fingerprinting program based on attention feature fusion, which, when executed by a processor, implements the federated learning radio frequency fingerprinting method based on attention feature fusion as described above.

[0015] To achieve the above objectives, a third aspect of the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described federated learning radio frequency fingerprint recognition method based on attention feature fusion.

[0016] To achieve the above objectives, a fourth aspect of this invention proposes a federated learning radio frequency fingerprinting system based on attention feature fusion, comprising: multiple clients and a server; each client acquires a corresponding radio frequency signal as a training dataset; each client trains a lightweight radio frequency fingerprinting model based on depthwise separable grouped convolution according to the corresponding training dataset to update local parameters, and sends the local parameters corresponding to each client to the server; the server performs weighted aggregation of the received local parameters according to the training sample size of each client, and sends the aggregated parameters as global parameters to each client for iterative training; wherein, in each round of iterative training, each client uses an attention feature fusion module to fuse the global parameters received in the current round and the local parameters updated in the previous round for personalized adjustments; the server uses an attention feature fusion module to fuse the weighted aggregated parameters and the global parameters from the previous round to obtain the global parameters for the current round.

[0017] In addition, the federated learning radio frequency fingerprint recognition system based on attention feature fusion proposed in the above embodiments of the present invention may also have the following additional technical features:

[0018] Optionally, the lightweight RF fingerprint recognition model includes an RNN denoising module, a one-dimensional convolutional layer, grouped convolutions, depthwise separable convolutions, a one-dimensional max pooling layer, and a fully connected layer.

[0019] Optionally, the lightweight RF fingerprint recognition model uses an RNN denoising module to denoise the input RF signal to obtain a denoised signal. The denoised signal is then processed by a one-dimensional convolutional layer, grouped convolution, and depthwise separable convolution to extract features, gradually reducing the spatial dimension of the features. Finally, the extracted features are mapped to the category dimension through a one-dimensional max pooling layer and a fully connected layer to construct the relationship between features and categories.

[0020] Optionally, the lightweight RF fingerprint recognition model uses the Adam optimizer and cross-entropy loss function to optimize parameters during training, with a learning rate of 0.001, momentum of 0.9, and 100 training iterations. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the federated learning radio frequency fingerprinting method based on attention feature fusion according to an embodiment of the present invention.

[0022] Figure 2 This is a diagram of the AFF-Fed algorithm framework according to an embodiment of the present invention;

[0023] Figure 3 This is a structural diagram of a lightweight radio frequency fingerprint recognition model based on depthwise separable grouped convolution according to an embodiment of the present invention;

[0024] Figure 4 The performance results of the DSGL_Net model on LoRa signals according to an embodiment of the present invention;

[0025] Figure 5 The convergence speed and stability analysis results of the DSGL_Net model according to an embodiment of the present invention are shown below;

[0026] Figure 6 Four different heterogeneous scenarios were set up on the LoRa dataset according to an embodiment of the present invention;

[0027] Figure 7 The sample distribution in four heterogeneous scenarios according to embodiments of the present invention;

[0028] Figure 8 The results represent the average recognition performance in four heterogeneous scenarios according to embodiments of the present invention. Detailed Implementation

[0029] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0030] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the invention to those skilled in the art.

[0031] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0032] refer to Figure 1 As shown, the federated learning radio frequency fingerprinting method based on attention feature fusion according to an embodiment of the present invention includes the following steps:

[0033] S101, multiple clients acquire their respective radio frequency signals as training datasets.

[0034] As an example, a wireless signal acquisition system can be built based on the software-defined radio platform GNU Radio and a USRP receiver to collect signals from ten types of LoRa wireless communication modules in actual communication scenarios. LoRa signals are collected using a USRP-LW X310 at a frequency of 433MHz, with a sampling frequency of 10MHz and a bandwidth of 1MHz. The number of samples for each type of LoRa signal collected is 2500, and the signal length is 6400.

[0035] S102, each client trains a lightweight RF fingerprint recognition model based on depthwise separable grouped convolution according to the corresponding training dataset to update local parameters, and sends the local parameters corresponding to each client to the server.

[0036] As an example, such as Figure 3 As shown, the lightweight RF fingerprint recognition model includes an RNN denoising module, a one-dimensional convolutional layer, grouped convolutions, depthwise separable convolutions, a one-dimensional max pooling layer, and a fully connected layer.

[0037] As an example, the lightweight RF fingerprint recognition model uses an RNN denoising module to denoise the input RF signal to obtain a denoised signal. The denoised signal is then processed through a one-dimensional convolutional layer, grouped convolution, and depthwise separable convolution to extract features, gradually reducing the spatial dimension of the features. Finally, the extracted features are mapped to the category dimension through a one-dimensional max pooling layer and a fully connected layer to construct the relationship between features and categories.

[0038] In other words, a lightweight recognition network model based on depthwise separable grouped convolutions is designed as the foundation model for the AFF-Fed algorithm, enabling its deployment on resource-constrained edge embedded devices as the basis for federated training. The lightweight model introduces a recurrent convolutional neural network module, which can learn the statistical characteristics of the input data and the statistical characteristics of noise, thereby achieving noise separation and removal. The lightweight model uses depthwise separable grouped convolutions for feature extraction, which reduces the number of model parameters and computational cost, while further improving the model's feature extraction capability, generalization ability, and regularization effect. In other words, it can enhance the model's feature extraction capability and reduce the risk of overfitting.

[0039] As a specific example, the LoRa dataset has 2000 training samples and 500 test samples. The lightweight RF fingerprint recognition model uses the Adam optimizer and cross-entropy loss function to optimize the parameters during training, with a learning rate of 0.001, momentum of 0.9, and 100 training iterations.

[0040] S103, the server performs weighted aggregation on the received local parameters based on the training sample size of each client, and sends the aggregated parameters as global parameters to each client for iterative training. In each round of iterative training, each client uses the attention feature fusion module to fuse the global parameters received in the current round with the local parameters updated in the previous round for personalized adjustments. The server uses the attention feature fusion module to fuse the weighted aggregated parameters with the global parameters from the previous round to obtain the global parameters for the current round.

[0041] In other words, a personalized federated recognition algorithm based on attention feature fusion is designed, introducing attention feature fusion modules on both the client and server sides to optimize federated training. Specifically, on the server side, the received local parameters are first weighted and aggregated according to the training sample size of each client. Then, the weighted aggregated parameters of the current round are fused with the global parameters of the previous round to obtain the global parameters of the current round. The global parameters are then sent to each client. On each client side, an attention feature fusion module is introduced to fuse the global parameters received in the current round with the local parameters updated in the previous training, so as to make personalized adjustments, reduce model drift, and adapt to the characteristics of the local dataset more quickly. Each client trains a lightweight RF fingerprint recognition model based on depthwise separable grouped convolution based on the fused parameters to update the local parameters, and sends the corresponding local parameters of each client to the server. This process is repeated until the training iterations are reached. In the detection phase, the client uses the local data for detection to obtain the recognition results.

[0042] As a specific example, the performance of the designed lightweight recognition model based on depthwise separable grouped convolutions is analyzed on the LoRa dataset. The performance results of the DSGL_Net model on LoRa signals are referenced. Figure 4 Experimental results show that the DSGL_Net model exhibits excellent recognition performance across various signal lengths, demonstrating strong feature extraction and generalization capabilities when recognizing LoRa signals. Furthermore, it achieves high accuracy while maintaining low parameter and computational complexity, meaning reduced hardware requirements for practical deployment, making it suitable for resource-constrained edge embedded devices, and offering shorter inference times. The convergence speed and stability analysis results of the DSGL_Net model are provided below. Figure 5 Experimental results show that the convergence curve of the DSGL_Net model has smaller fluctuations compared to other comparison networks, and the number of iterations required to achieve convergence is also smaller. This indicates that the DSGL_Net model has higher stability and faster convergence speed during training.

[0043] The performance of the designed attention-based feature fusion-based personalized federated radio frequency fingerprinting (AFF-Fed) algorithm was analyzed on a set LoRa heterogeneous dataset. The AFF-Fed algorithm framework is referenced from... Figure 2 Four different heterogeneous scene references were set up on the LoRa dataset. Figure 6 The sample distribution in the four scenarios is referenced. Figure 7 The average recognition performance results in four heterogeneous scenarios are referenced. Figure 8Experimental results show that, in four heterogeneous scenes of the LoRa dataset, the AFF-Fed algorithm outperforms other compared federated learning algorithms in terms of recognition performance. Furthermore, the AFF-Fed algorithm exhibits faster convergence speed and better stability.

[0044] In summary, the federated learning RFID fingerprinting method based on attentional feature fusion according to embodiments of the present invention first designs a lightweight recognition model (DSGL_Net) based on depthwise separable grouped convolution as the base model. Then, by introducing an Attentional Feature Fusion (AFF) module on the server side, the weighted aggregated parameters of the current round and the global parameters of the previous round are fused to obtain the global parameters of the current round. This dynamic fusion mechanism can utilize historical information to smooth model updates, reduce fluctuations caused by random client selection, thereby accelerating convergence and improving model stability. Simultaneously, it can also alleviate the impact of uneven data distribution on model performance, thereby improving the model's generalization ability. Similarly, the AFF-Fed algorithm introduces an AFF module on the client side, fusing the global model received in the current round and the local model updated in the previous training to achieve personalized adjustments, enabling the personalized adjusted model to adapt more quickly to the characteristics of the local dataset. This alleviates the heterogeneity between datasets of various clients in distributed scenarios, and the model after federated training convergence has a high RFID fingerprinting accuracy, making it suitable for deployment in edge distributed scenarios.

[0045] In addition, the present invention also proposes a computer-readable storage medium storing a federated learning radio frequency fingerprinting program based on attention feature fusion, which, when executed by a processor, implements the federated learning radio frequency fingerprinting method based on attention feature fusion as described above.

[0046] In addition, this invention also proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described federated learning radio frequency fingerprint recognition method based on attention feature fusion.

[0047] Figure 2 This is a block diagram of a federated learning radio frequency fingerprint recognition system based on attention feature fusion according to an embodiment of the present invention. Figure 2 As shown, the federated learning radio frequency fingerprint recognition system includes: multiple clients 10 and a server 20;

[0048] In this process, multiple clients 10 acquire corresponding radio frequency signals as training datasets. Each client trains a lightweight radio frequency fingerprint recognition model based on depthwise separable grouped convolutions according to its corresponding training dataset to update local parameters and sends the local parameters corresponding to each client to the server. The server 20 performs weighted aggregation on the received local parameters according to the training sample size of each client and sends the aggregated parameters as global parameters to each client for iterative training. In each round of iterative training, each client uses an attention feature fusion module to fuse the global parameters received in the current round and the local parameters updated in the previous round for personalized adjustments. The server uses an attention feature fusion module to fuse the weighted aggregated parameters and the global parameters from the previous round to obtain the global parameters for the current round.

[0049] As an example, the lightweight RF fingerprint recognition model includes an RNN denoising module, a one-dimensional convolutional layer, grouped convolutions, depthwise separable convolutions, a one-dimensional max pooling layer, and a fully connected layer.

[0050] As an example, the lightweight RF fingerprint recognition model uses an RNN denoising module to denoise the input RF signal to obtain a denoised signal. The denoised signal is then processed through a one-dimensional convolutional layer, grouped convolution, and depthwise separable convolution to extract features, gradually reducing the spatial dimension of the features. Finally, the extracted features are mapped to the category dimension through a one-dimensional max pooling layer and a fully connected layer to construct the relationship between features and categories.

[0051] As an example, the lightweight RF fingerprint recognition model uses the Adam optimizer and cross-entropy loss function to optimize parameters during training, with a learning rate of 0.001, momentum of 0.9, and 100 training iterations.

[0052] It should be noted that the foregoing explanation of the embodiments of the federated learning RF fingerprinting method based on attention feature fusion also applies to the federated learning RF fingerprinting system based on attention feature fusion in this embodiment, and will not be repeated here.

[0053] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0055] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0056] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0057] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0058] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0059] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0060] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0061] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0062] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0063] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms should not be construed as necessarily referring to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0064] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A federated learning radio frequency fingerprint recognition method based on attention feature fusion, characterized in that, Includes the following steps: Multiple clients acquire their respective radio frequency signals as training datasets; Each client trains a lightweight RF fingerprint recognition model based on depthwise separable grouped convolution according to the corresponding training dataset to update local parameters and sends the local parameters corresponding to each client to the server. The server performs weighted aggregation on the received local parameters based on the training sample size of each client, and sends the aggregated parameters as global parameters to each client for iterative training. In each round of iterative training, each client uses the attention feature fusion module to fuse the global parameters received in the current round with the local parameters updated in the previous round for personalized adjustments. The server uses the attention feature fusion module to fuse the weighted aggregated parameters with the global parameters from the previous round to obtain the global parameters for the current round.

2. The federated learning RFID fingerprinting method based on attention feature fusion as described in claim 1, characterized in that, The lightweight RF fingerprint recognition model includes an RNN denoising module, a one-dimensional convolutional layer, grouped convolutions, depthwise separable convolutions, a one-dimensional max pooling layer, and a fully connected layer.

3. The federated learning RFID fingerprinting method based on attention feature fusion as described in claim 2, characterized in that, The lightweight RF fingerprint recognition model uses an RNN denoising module to denoise the input RF signal to obtain a denoised signal. The denoised signal is then processed through a one-dimensional convolutional layer, grouped convolution, and depthwise separable convolution to extract features, gradually reducing the spatial dimension of the features. Finally, the extracted features are mapped to the category dimension through a one-dimensional max pooling layer and a fully connected layer to construct the relationship between features and categories.

4. The federated learning RFID fingerprinting method based on attention feature fusion as described in claim 3, characterized in that, The lightweight RF fingerprint recognition model uses the Adam optimizer and cross-entropy loss function to optimize parameters during training, with a learning rate of 0.001, momentum of 0.9, and 100 training iterations.

5. A computer-readable storage medium, characterized in that, It stores a federated learning radio frequency fingerprinting program based on attention feature fusion, which, when executed by a processor, implements the federated learning radio frequency fingerprinting method based on attention feature fusion as described in any one of claims 1-4.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the federated learning radio frequency fingerprinting method based on attention feature fusion as described in any one of claims 1-4.

7. A federated learning radio frequency fingerprint recognition system based on attention feature fusion, characterized in that, include: Multiple clients and servers are involved, with each client acquiring its corresponding radio frequency signal as a training dataset. Each client trains a lightweight RFID fingerprinting model based on depthwise separable grouped convolutions using its corresponding training dataset to update its local parameters and sends these local parameters to the server. The server then weights and aggregates the received local parameters based on the number of training samples from each client and sends the aggregated parameters as global parameters to each client for iterative training. In each iteration, each client uses an attention feature fusion module to fuse the global parameters received in the current iteration with the updated local parameters from the previous iteration for personalized adjustments. The server uses the attention feature fusion module to fuse the weighted aggregated parameters with the global parameters from the previous iteration to obtain the global parameters for the current iteration.

8. The federated learning radio frequency fingerprint recognition system based on attention feature fusion as described in claim 7, characterized in that, The lightweight RF fingerprint recognition model includes an RNN denoising module, a one-dimensional convolutional layer, grouped convolutions, depthwise separable convolutions, a one-dimensional max pooling layer, and a fully connected layer.

9. The federated learning radio frequency fingerprint recognition system based on attention feature fusion as described in claim 8, characterized in that, The lightweight RF fingerprint recognition model uses an RNN denoising module to denoise the input RF signal to obtain a denoised signal. The denoised signal is then processed through a one-dimensional convolutional layer, grouped convolution, and depthwise separable convolution to extract features, gradually reducing the spatial dimension of the features. Finally, the extracted features are mapped to the category dimension through a one-dimensional max pooling layer and a fully connected layer to construct the relationship between features and categories.

10. The federated learning radio frequency fingerprint recognition system based on attention feature fusion as described in claim 9, characterized in that, The lightweight RF fingerprint recognition model uses the Adam optimizer and cross-entropy loss function to optimize parameters during training, with a learning rate of 0.001, momentum of 0.9, and 100 training iterations.

Images