Method, apparatus, device, medium and product for spectrum sensing against malicious user attacks

Abstract

The application discloses a spectrum sensing method and device for preventing malicious user attacks, equipment, medium and product. The method comprises the following steps: taking a secondary user participating in a spectrum sensing task as a node, and taking a cooperative sensing relationship as an edge to construct a graph structure; extracting node structure data of a to-be-identified secondary user from the graph structure, reconstructing the node structure data by using a trained graph neural network, and obtaining reconstructed node structure data of the to-be-identified secondary user; obtaining a reconstruction error of the to-be-identified secondary user according to the difference between the node structure data and the reconstructed node structure data; identifying a malicious secondary user from the to-be-identified secondary user according to the reconstruction error; and performing cooperative spectrum sensing by using non-malicious secondary users. The embodiment of the application can prevent the malicious secondary user from sending an incorrect sensing result to affect the final decision of the fusion center, and improve the cooperative spectrum sensing performance and the result reliability.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a spectrum sensing method, apparatus, device, medium and product for preventing malicious user attacks. Background Technology

[0002] Spectrum sensing, a core technology of cognitive radio (CR), enables spectrum sharing by detecting idle frequency bands and allowing secondary users to access these bands. In existing centralized collaborative spectrum sensing architectures, a fusion center is responsible for issuing spectrum sensing tasks. Multiple secondary users (SUs) independently complete their tasks and then upload their results to the fusion center. The fusion center aggregates and processes all received results to ultimately determine the existence of a primary user (PU). However, collaborative sensing is vulnerable to attacks from malicious users (i.e., malicious secondary users): these malicious users impersonate honest nodes and send incorrect sensing results to the fusion center, misleading its decisions and severely damaging the sensing performance and reliability of the cognitive radio network. Summary of the Invention

[0003] This application provides a spectrum sensing method, apparatus, device, medium, and product for preventing malicious user attacks, in order to solve the problem that the prior art is vulnerable to malicious user attacks.

[0004] To achieve the above objectives, embodiments of this application provide a spectrum sensing method for preventing malicious user attacks, including: A graph structure is constructed using secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges; The node structure data of the secondary user to be identified is extracted from the graph structure, and the node structure data is reconstructed using a trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified. The graph neural network is trained by a sample graph structure constructed with sample secondary users as nodes. Its training objective is to learn node embedding representations and edge embedding representations, which are used to identify malicious secondary users. Based on the difference between the node structure data and the reconstructed node structure data, the reconstruction error of the sub-user to be identified is obtained; Based on the reconstruction error, identify malicious secondary users from the users to be identified; Collaborative spectrum sensing is conducted using non-malicious secondary users.

[0005] As an improvement to the above scheme, the step of identifying malicious secondary users from the users to be identified based on the reconstruction error includes: Using the reconstruction error, the secondary users to be identified that meet the preset reconstruction error conditions are identified as candidate malicious secondary users; The global spectrum sensing result is obtained based on the local spectrum sensing result of the sub-user to be identified and the local spectrum sensing results of all sub-users participating in the same spectrum sensing task as the sub-user to be identified. The perception error of the candidate malicious secondary user is obtained based on the difference between the local spectrum perception result and the global spectrum perception result. Based on the perceived error, malicious secondary users are identified from the candidate malicious secondary users.

[0006] As an improvement to the above scheme, the step of identifying malicious secondary users from candidate malicious secondary users based on the perceived error includes: Obtain the similarity between the perceived error and the reconstruction error; When the similarity is less than a preset similarity threshold, obtain the distribution of neighboring candidate malicious secondary users of the candidate malicious secondary user corresponding to the candidate malicious secondary user whose similarity is less than the preset similarity threshold; Candidate malicious secondary users whose distribution satisfies the preset distribution are identified as malicious secondary users.

[0007] As an improvement to the above scheme, the step of using the reconstruction error to identify secondary users who meet the preset reconstruction error conditions as candidate malicious secondary users includes: Obtain the normalized value of the reconstruction error; When the normalized value of the reconstruction error is greater than the preset reconstruction error threshold, the user to be identified corresponding to the error threshold is identified as a candidate malicious secondary user.

[0008] As an improvement to the above solution, obtaining the similarity between the perceptual error and the reconstruction error includes: Obtain the normalized value of the perception error; The similarity is obtained by using the normalized value of the reconstruction error and the normalized value of the perception error.

[0009] As an improvement to the above scheme, the step of extracting the node structure data of the secondary user to be identified from the graph structure, and reconstructing the node structure data using a trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified, includes: Extract the node structure data of the sub-user to be identified from the graph structure; The attention score between the user to be identified and its neighboring users is calculated using a trained graph neural network through an attention mechanism. The attention scores are normalized to obtain the attention weights; Using the attention weights, the embedded representation of the node structure data is obtained; The embedded representation of the node structure data is reconstructed by the encoder to obtain the reconstructed node structure data.

[0010] To achieve the above objectives, embodiments of this application also provide a spectrum sensing device for preventing malicious user attacks, comprising: The module is used to construct a graph structure with secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges. The reconstruction module is used to extract the node structure data of the secondary user to be identified from the graph structure, and reconstruct the node structure data using a trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified; wherein, the graph neural network is trained by a sample graph structure constructed with sample secondary users as nodes, and its training objective is to learn node embedding representations and edge embedding representations, which are used to identify malicious secondary users. The calculation module is used to obtain the reconstruction error of the sub-user to be identified based on the difference between the node structure data and the reconstructed node structure data. The identification module is used to identify malicious secondary users from the users to be identified based on the reconstruction error. The perception module is used to conduct collaborative spectrum sensing using non-malicious secondary users.

[0011] To achieve the above objectives, this application also provides a spectrum sensing device for preventing malicious user attacks, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the spectrum sensing method for preventing malicious user attacks as described above.

[0012] To achieve the above objectives, embodiments of this application also provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program; wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the spectrum sensing method for preventing malicious user attacks as described above.

[0013] To achieve the above objectives, embodiments of this application also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the spectrum sensing method for preventing malicious user attacks as described above.

[0014] Compared with existing technologies, the spectrum sensing method, apparatus, device, medium, and product for preventing malicious user attacks provided in this application embodiment constructs a graph structure using secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges; extracts node structure data of the secondary user to be identified from the graph structure, and reconstructs the node structure data using a trained graph neural network to obtain reconstructed node structure data of the secondary user to be identified; wherein, the graph neural network is trained by a sample graph structure constructed with sample secondary users as nodes, and its training objective is to learn node embedding representations and edge embedding representations, which are used to identify malicious secondary users; the reconstruction error of the secondary user to be identified is obtained based on the difference between the node structure data and the reconstructed node structure data; malicious secondary users are identified from the users to be identified based on the reconstruction error; and collaborative spectrum sensing is performed using non-malicious secondary users. Therefore, this application's embodiment utilizes graph neural networks to reconstruct the node structure data of the secondary users to be identified during cooperative spectrum sensing. By identifying malicious secondary users based on their reconstruction errors, it prevents malicious secondary users from sending erroneous sensing results that could influence the final decision of the fusion center, significantly improving the performance and reliability of multi-user cooperative spectrum sensing in cognitive radio networks. Furthermore, the use of graph neural networks can accurately capture and transmit the correlation information between secondary users, thereby enhancing the performance of detecting malicious secondary users. The entire cooperative spectrum sensing process does not increase the computational and network load of spectrum sensing, nor does it incur additional performance overhead. Attached Figure Description

[0015] Figure 1 This is a flowchart of a spectrum sensing method for preventing malicious user attacks provided in an embodiment of this application; Figure 2 This is an architecture diagram of a collaborative sensing system provided in an embodiment of this application; Figure 3 This is a structural block diagram of a spectrum sensing device for preventing malicious user attacks provided in an embodiment of this application; Figure 4 This is a structural block diagram of a spectrum sensing device for preventing malicious user attacks provided in an embodiment of this application. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0017] In the description of this application, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0018] In this application description, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0019] In this application description, the terms "first," "second," etc., 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, for example, in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus 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 apparatus. The term "based on" means "at least partially based on." The term "according to" means "at least partially according to." The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments." The term "and / or" means at least one of the connected objects, such as A and / or B, indicating three cases: including only A, only B, and both A and B. Unless otherwise stated, the term "multiple" means two or more.

[0020] It's worth noting that a graph structure is a data structure composed of nodes (also called vertices) and edges, which can naturally represent various complex relationships and networks, such as social networks, transportation networks, and molecular structures. In a graph structure, nodes represent entities, and edges represent the relationships between entities. Graph Neural Networks (GNNs) are deep learning models specifically designed to process graph structures.

[0021] The basic idea of ​​graph neural networks is to obtain the embedding representation (i.e., feature vector representation) of a target node by capturing its neighboring nodes. Neighboring nodes refer to other nodes directly connected to the target node. The embedding representation of the target node is updated by aggregating the feature information of its neighbors. Specifically, the basic principles of graph neural networks include the following steps: Information aggregation is a crucial step in graph neural network architecture. It integrates information from each node and its neighbors to update the embedding representation of each node. Information aggregation methods such as graph convolution, pooling operations, and attention mechanisms enable graph neural networks to delve into complex patterns within graphs, capture subtle connections between nodes, and provide powerful support for various graph-related tasks. In this way, graph neural networks continuously optimize their node representation updates, gradually enriching and refining their multi-layered network structures, demonstrating a profound understanding of graph data and powerful processing capabilities.

[0022] Information transfer, a crucial step following information aggregation, is responsible for passing the updated node embeddings to neighboring nodes to further refine the representation of each node. For example, feature vectors can be directly concatenated to form a rich information set, or the dynamic updates of recurrent neural networks can be used to capture the temporal nature of information transfer. Alternatively, Long Short-Term Memory (LSTM) networks provide a powerful memory mechanism for graph neural networks, enabling them to learn and remember key information during information transfer, thereby capturing more subtle and profound structural features in complex graph networks. Through these methods, graph neural networks not only enhance information exchange between nodes but also improve the overall network's deep understanding of graph structures, providing more accurate and comprehensive representation capabilities for various graph-related tasks.

[0023] Graph neural networks (GNNs) offer excellent flexibility and generalization capabilities, enabling them to handle various types of graph data, such as directed graphs, undirected graphs, and weighted graphs. Furthermore, GNNs can incorporate multiple feature information from nodes and edges to enhance model performance.

[0024] It is worth noting that the collaborative awareness system architecture is as follows: Figure 2 As shown, PU represents the primary user, FC represents the fusion center, SU represents the secondary user in the collaborative perception process, and MU represents the malicious secondary user in the collaborative perception process. Malicious secondary users are a special type of secondary user.

[0025] Each participating secondary user sends its local sensing results to the fusion center. The fusion center then makes a decision on the received local sensing results according to a preset fusion criterion. The decision method is divided into hard decision (also known as voting fusion criterion) and soft decision. Hard decision means that each secondary user sends "0" or "1" information representing the existence of the primary user to the fusion center. The fusion center selects an appropriate fusion criterion for merging based on system requirements. The AND fusion criterion, OR fusion criterion, and KN fusion criterion are currently the most commonly used hard decision criteria; the OR criterion (K=1) and the AND criterion (K=N) can be considered special forms of the KN fusion criterion. Soft decision means that each secondary user sends its test statistic or the raw sensing data to the fusion center. The fusion center selects an appropriate fusion criterion to make the final decision. The likelihood ratio algorithm and the equal gain algorithm are the most commonly used soft decision cooperative spectrum sensing algorithms. For example, when using the KN fusion criterion to determine the existence of the primary user, a threshold K needs to be set. When more than or equal to K local sensing results believe that the primary user exists, the fusion center's final decision is that the primary user exists; otherwise, the primary user does not exist.

[0026] Secondary users employ spectrum sensing methods including energy detection, matched filtering-based detection, and cyclostationary feature detection to obtain local sensing results. For example, after the fusion center issues a spectrum sensing task (i.e., a collaborative spectrum sensing task), each secondary user obtains a detection probability using the energy detection method within a certain time period. Assuming a threshold of 0.8, a detection probability exceeding 0.8 indicates the presence of the primary user, while a probability less than 0.8 indicates the absence of the primary user. Each secondary user then sends its local sensing results to the fusion center.

[0027] Define the set of secondary users Where M represents the total number of users, Indicates the first Each user.

[0028] The local perception result of the i-th sub-user in centralized collaborative perception Represented as: ,in, This represents the total number of spectrum sensing tasks. Indicates the first The user's perception Local awareness results for each task , Indicates the first The user's perception The local awareness result for this task indicates that the primary user does not exist. Indicates the first The user's perception The local perception results for each task exist primarily for the user.

[0029] In centralized collaborative sensing, the global sensing result of the fusion center is represented as follows: ,in The ruling of the Fusion Center indicates The global perception results for each task , The ruling of the Fusion Center indicates The global awareness result for this task indicates that the primary user does not exist. The ruling of the Fusion Center indicates The global perception results for each task exist primarily for the user.

[0030] See Figure 1 , Figure 1 This is a flowchart of a spectrum sensing method for preventing malicious user attacks provided in an embodiment of this application. The spectrum sensing method for preventing malicious user attacks includes: S1. Construct a graph structure with secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges; It is worth noting that in this embodiment, the fusion center leads the entire collaborative spectrum sensing process. A graph structure is constructed using secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges. For example, if two secondary users jointly participate in the same spectrum sensing task (i.e., jointly participate in collaborative sensing of the same spectrum segment), then there is an undirected edge between them, thus constructing the graph structure. Secondary users within the graph structure are divided into non-malicious and malicious secondary users. Malicious secondary users are then identified using reconstruction errors, so that when spectrum sensing is needed again, non-malicious secondary users can be used for collaborative spectrum sensing.

[0031] For example, the fusion center broadcasts the same spectrum sensing task to all secondary users within the sensing area. After receiving the task, the secondary users participate in completing the spectrum sensing task and send their local spectrum sensing results to the fusion center. Among these secondary users participating in the same spectrum sensing task, there may be malicious secondary users. Malicious secondary users can be identified by utilizing reconstruction errors, so that non-malicious secondary users can be used for collaborative spectrum sensing in the future, without the participation of malicious secondary users.

[0032] S2. Extract the node structure data of the secondary user to be identified from the graph structure, and reconstruct the node structure data using the trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified; wherein, the graph neural network is trained by a sample graph structure constructed with sample secondary users as nodes, and its training objective is to learn node embedding representation and edge embedding representation, which are used to identify malicious secondary users. It is worth noting that, in this embodiment, a sample graph structure is pre-constructed using sample secondary users as nodes. Using this sample graph structure, the training objective is to learn node embedding representations (i.e., the embedding representations of nodes) and edge embedding representations (i.e., the embedding representations of edges) to complete the training of the graph neural network, thereby achieving malicious user identification based on the graph structure. Here, sample secondary users include sample non-malicious secondary users and sample malicious secondary users. Specifically, the training objective is to learn the node embedding representations and edge embedding representations of sample non-malicious secondary users. Therefore, the greater the difference between the node structure data of the user to be identified and the reconstructed node structure data, the more likely the user to be identified as a malicious secondary user.

[0033] The specific training process includes information aggregation and information transmission, which will not be elaborated here.

[0034] The node structure data of the secondary users to be identified is essentially a matrix describing the topological connections / structural features between nodes in the graph structure, including: the adjacency matrix of the secondary users to be identified and / or the feature matrix of the secondary users to be identified. The adjacency matrix reflects the description of the connections between nodes, while the feature matrix reflects the feature description of each node.

[0035] S3. Based on the difference between the node structure data and the reconstructed node structure data, obtain the reconstruction error of the sub-user to be identified; It's worth noting that reconstruction error represents the difference between the node structure data of the user to be identified and the reconstructed node structure data, reflecting the degree of deviation between the user to be identified and a non-malicious user. The greater the difference between the node structure data of the user to be identified and the reconstructed node structure data, the greater the reconstruction error, and the more likely the user to be identified as a malicious user. Therefore, reconstruction error can also be considered as identification error.

[0036] S4. Based on the reconstruction error, identify malicious secondary users from the users to be identified; It is worth noting that this application utilizes reconstruction error to identify malicious secondary users from among the users to be identified, effectively distinguishing between non-malicious and malicious secondary users and significantly improving the accuracy of malicious secondary user identification. For example, the reconstruction error of a certain secondary user to be identified is normalized to obtain a normalized value of the reconstruction error. When this normalized value is greater than a preset reconstruction error threshold, the secondary user to be identified is identified as a malicious secondary user.

[0037] S5. Utilize non-malicious secondary users for collaborative spectrum sensing.

[0038] It is worth noting that, in this embodiment, after identifying a malicious secondary user, cooperative spectrum sensing is subsequently performed using non-malicious secondary users. Specifically, during subsequent spectrum sensing, the fusion center broadcasts spectrum sensing tasks to non-malicious secondary users. Upon receiving the task, the non-malicious secondary users perform spectrum sensing and return their local sensing results to the fusion center for final judgment. This prevents malicious secondary users from sending erroneous sensing results that could influence the fusion center's final decision, significantly improving the performance and reliability of multi-user cooperative spectrum sensing in cognitive radio networks. Furthermore, this embodiment employs graph neural networks, which can accurately capture and transmit correlation information between secondary users, thereby enhancing the performance of detecting malicious secondary users. The entire cooperative spectrum sensing process does not increase the computational and network load of spectrum sensing, nor does it incur additional performance overhead.

[0039] In one optional embodiment, identifying malicious secondary users from the users to be identified based on the reconstruction error includes: Using the reconstruction error, the secondary users to be identified that meet the preset reconstruction error conditions are identified as candidate malicious secondary users; The global spectrum sensing result is obtained based on the local spectrum sensing result of the sub-user to be identified and the local spectrum sensing results of all sub-users participating in the same spectrum sensing task as the sub-user to be identified. The perception error of the candidate malicious secondary user is obtained based on the difference between the local spectrum perception result and the global spectrum perception result. Based on the perceived error, malicious secondary users are identified from the candidate malicious secondary users.

[0040] It is worth noting that, in order to further improve the accuracy of identification, this application embodiment combines reconstruction error and perception error to identify malicious secondary users. This can effectively avoid the limitations of single-dimensional error judgment, significantly improve the robustness and anti-interference ability of malicious secondary user identification, and thus ensure the perception performance and reliability of the cognitive radio network.

[0041] For example, the fusion center broadcasts the same spectrum sensing task to the sub-user to be identified and other sub-users in the same area as the sub-user to be identified. After receiving the task, these sub-users (including the sub-user to be identified and other sub-users in the same area) will participate in completing the spectrum sensing task and send their respective local spectrum sensing results to the fusion center.

[0042] The fusion center, according to preset fusion criteria, makes a judgment on the received local spectrum sensing results (i.e., the local spectrum sensing results of secondary users participating in the same spectrum sensing task, including: the local spectrum sensing results of the secondary user to be identified and the local spectrum sensing results of other secondary users in the same area) to obtain the global spectrum sensing results. This allows the center to identify malicious secondary users from among the candidate malicious secondary users by utilizing their sensing errors. For example, if the normalized value of the sensing error of a candidate malicious secondary user is greater than a preset sensing error threshold, that candidate malicious secondary user is identified as a malicious secondary user.

[0043] In an optional embodiment, the step of using the reconstruction error to identify secondary users who meet the preset reconstruction error conditions as candidate malicious secondary users includes: Obtain the normalized value of the reconstruction error; When the normalized value of the reconstruction error is greater than the preset reconstruction error threshold, the user to be identified corresponding to the error threshold is identified as a candidate malicious secondary user.

[0044] This application's embodiments quantify the abnormality of the secondary users to be identified by normalizing the reconstruction error. The normalized value of the reconstruction error can more intuitively reflect the degree of deviation between the secondary user to be identified and a non-malicious secondary user, and is used as an important basis for initially judging whether the secondary user to be identified is a malicious secondary user. Here, secondary users initially judged to be malicious secondary users are called candidate malicious secondary users. By setting a certain reconstruction error threshold, malicious secondary users who may pose a threat to the network can be quickly identified, which not only improves the accuracy of malicious secondary user detection, but also provides strong protection for the secure operation of the network.

[0045] Specifically, the sub-user to be identified is calculated according to the following formula. Reconstruction error :

[0046] in, For the secondary user to be identified The node structure data, It is the reconstructed node structure data of the secondary user to be identified; The normalized value of the reconstruction error of the sub-user to be identified is calculated according to the following formula. :

[0047] in, Indicates the secondary user to be identified Reconstruction error.

[0048] In an optional embodiment, calculating the similarity between the perceived error and the reconstruction error includes: Obtain the normalized value of the perception error; The similarity is obtained by using the normalized value of the reconstruction error and the normalized value of the perception error.

[0049] It is worth noting that the normalized value of the reconstruction error is positively correlated with the degree of similarity, and the normalized value of the perception error is also positively correlated with the degree of similarity.

[0050] Specifically, the perceived error of candidate malicious secondary users within a preset time period is calculated according to the following formula. :

[0051] in, This represents the local perception result of candidate malicious secondary user i on spectrum perception task n. This represents the global perception result of the fusion center for spectrum sensing task n, where N represents the number of spectrum sensing tasks within a preset time period.

[0052] The normalized value of the perceived error is calculated using the following formula. :

[0053] The similarity is calculated using the following formula. :

[0054] Of course, the calculation of similarity is not limited to the above formula; it only needs to satisfy the normalized value of the reconstruction error. The normalized value of the perceptual error is inversely correlated with the degree of similarity. It should be inversely correlated with the degree of similarity.

[0055] In an optional embodiment, identifying malicious secondary users from candidate malicious secondary users by utilizing the similarity between the perception error and the reconstruction error includes: Obtain the similarity between the perceived error and the reconstruction error; When the similarity is less than a preset similarity threshold, obtain the distribution of neighboring candidate malicious secondary users of the candidate malicious secondary user corresponding to the candidate malicious secondary user whose similarity is less than the preset similarity threshold; Candidate malicious secondary users whose distribution satisfies the preset distribution are identified as malicious secondary users.

[0056] In this embodiment, when the similarity between the perception error and the reconstruction error is less than a preset similarity threshold, it indicates a possible misidentification. This could be because the distribution of neighboring secondary users of a candidate malicious secondary user affects the perception of that user. This embodiment analyzes the distribution of neighboring candidate malicious secondary users when the similarity between the perception error and reconstruction error of a candidate malicious secondary user is less than the preset similarity threshold. Candidate malicious secondary users meeting the preset distribution criteria are then identified as malicious secondary users. This embodiment fully considers the influence of environmental factors and neighboring secondary users, thereby effectively improving the accuracy of the final decision made by the fusion center.

[0057] Optionally, the preset distribution includes: the total number of neighboring candidate malicious secondary users is less than a preset number.

[0058] It's worth noting that when the total number of neighboring candidate malicious users of a candidate malicious user is too large, it indicates that the candidate malicious user may have been influenced by neighboring candidate malicious users, and therefore, the candidate malicious user is identified as a non-malicious user. Conversely, when the total number of neighboring candidate malicious users of a candidate malicious user is too small, it indicates that the candidate malicious user was not influenced by neighboring candidate malicious users, and therefore, the candidate malicious user is identified as the final malicious user. Here, the neighboring candidate malicious users of a candidate malicious user refer to candidate malicious users directly connected to that candidate malicious user. Specifically, the preset number = M represents the total number of sub-users with the same spectrum sensing task.

[0059] In an optional embodiment, the step of extracting node structure data of the secondary user to be identified from the graph structure, and reconstructing the node structure data using a trained graph neural network to obtain reconstructed node structure data of the secondary user to be identified, includes: Extract the node structure data of the sub-user to be identified from the graph structure; The attention score between the user to be identified and its neighboring users is calculated using a trained graph neural network through an attention mechanism. The attention scores are normalized to obtain the attention weights; Using the attention weights, the embedded representation of the node structure data is obtained; The embedded representation of the node structure data is reconstructed by the encoder to obtain the reconstructed node structure data.

[0060] In one specific embodiment, the graph neural network calculates the sub-user to be identified through an attention mechanism. With neighboring secondary users Attention scores between (That is, the importance of the secondary user to be identified to its neighboring secondary users), the formula is as follows:

[0061] in, and Represents shared weight. and Used to distinguish different weight representations Indicates transpose. This represents the activation function. The graph attention mechanism computation function is represented. This represents the initial feature embedding vector of sub-user i (i.e., the initial feature representation of node i). This represents the initial feature embedding vector of subuser j (i.e., the initial feature representation of node j).

[0062] pass The function normalizes the attention scores to obtain the attention weights. The formula is as follows:

[0063] in, Let represent the set of neighboring secondary users of secondary user i, consisting of the neighboring nodes of secondary user i. Indicates secondary user With neighboring secondary users Attention scores between Indicates secondary user With neighboring secondary users Attention scores between them.

[0064] By utilizing attention weights, a new embedding representation of the node structure data is obtained. The formula is as follows:

[0065] in, Let i represent the set of neighboring secondary users. This represents the attention weight.

[0066] pass The function retrieves the embedded representation of the node structure data as the output reconstructed by the encoder, as shown in the following formula:

[0067] In the formula, T represents transpose. An embedded representation of node structure data.

[0068] See Figure 3 , Figure 3This is a structural block diagram of a spectrum sensing device 10 for preventing malicious user attacks provided in an embodiment of this application. The spectrum sensing device 10 for preventing malicious user attacks includes: Module 11 is used to construct a graph structure with secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges. The reconstruction module 12 is used to extract the node structure data of the secondary user to be identified from the graph structure, and reconstruct the node structure data using a trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified; wherein, the graph neural network is trained by a sample graph structure constructed with sample secondary users as nodes, and its training objective is to learn node embedding representation and edge embedding representation, which are used to identify malicious secondary users. Calculation module 13 is used to obtain the reconstruction error of the sub-user to be identified based on the difference between the node structure data and the reconstructed node structure data; The identification module 14 is used to identify malicious secondary users from the users to be identified based on the reconstruction error. The perception module 15 is used to perform cooperative spectrum perception using non-malicious secondary users.

[0069] Optionally, identifying malicious secondary users from the users to be identified based on the reconstruction error includes: Using the reconstruction error, the secondary users to be identified that meet the preset reconstruction error conditions are identified as candidate malicious secondary users; The global spectrum sensing result is obtained based on the local spectrum sensing result of the sub-user to be identified and the local spectrum sensing results of all sub-users participating in the same spectrum sensing task as the sub-user to be identified. The perception error of the candidate malicious secondary user is obtained based on the difference between the local spectrum perception result and the global spectrum perception result. Based on the perceived error, malicious secondary users are identified from the candidate malicious secondary users.

[0070] Optionally, identifying malicious secondary users from candidate malicious secondary users based on the perceived error includes: Obtain the similarity between the perceived error and the reconstruction error; When the similarity is less than a preset similarity threshold, obtain the distribution of neighboring candidate malicious secondary users of the candidate malicious secondary user corresponding to the candidate malicious secondary user whose similarity is less than the preset similarity threshold; Candidate malicious secondary users whose distribution satisfies the preset distribution are identified as malicious secondary users.

[0071] Optionally, the step of using the reconstruction error to identify secondary users who meet the preset reconstruction error conditions as candidate malicious secondary users includes: Obtain the normalized value of the reconstruction error; When the normalized value of the reconstruction error is greater than the preset reconstruction error threshold, the user to be identified corresponding to the error threshold is identified as a candidate malicious secondary user.

[0072] Optionally, obtaining the similarity between the perceptual error and the reconstruction error includes: Obtain the normalized value of the perception error; The similarity is obtained by using the normalized value of the reconstruction error and the normalized value of the perception error.

[0073] Optionally, the step of extracting the node structure data of the secondary user to be identified from the graph structure, and reconstructing the node structure data using a trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified, includes: Extract the node structure data of the sub-user to be identified from the graph structure; The attention score between the user to be identified and its neighboring users is calculated using a trained graph neural network through an attention mechanism. The attention scores are normalized to obtain the attention weights; Using the attention weights, the embedded representation of the node structure data is obtained; The embedded representation of the node structure data is reconstructed by the encoder to obtain the reconstructed node structure data.

[0074] It is worth noting that the working process of each module in the spectrum sensing device 10 for preventing malicious user attacks described in this application embodiment can refer to the working process of the spectrum sensing method for preventing malicious user attacks described in the above embodiment, and will not be repeated here.

[0075] This application provides a spectrum sensing device 10 for preventing malicious user attacks. It constructs a graph structure using secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges. The device extracts node structure data of the secondary user to be identified from the graph structure and reconstructs this data using a trained graph neural network to obtain reconstructed node structure data. The graph neural network is trained using a sample graph structure constructed with sample secondary users as nodes. Its training objective is to learn node embedding representations and edge embedding representations, which are used to identify malicious secondary users. The device obtains the reconstruction error of the secondary user to be identified based on the difference between the node structure data and the reconstructed node structure data. Based on the reconstruction error, malicious secondary users are identified from the users to be identified. Collaborative spectrum sensing is then performed using non-malicious secondary users. Therefore, this application's embodiment utilizes a graph neural network to reconstruct the node structure data of the secondary user to be identified during collaborative spectrum sensing, and then identifies malicious secondary users based on their reconstruction error. This prevents malicious secondary users from sending erroneous sensing results that could affect the final decision of the fusion center, significantly improving the performance and reliability of multi-user collaborative spectrum sensing in cognitive radio networks. Furthermore, the use of graph neural networks can accurately capture and transmit the correlation information between secondary users, thereby enhancing the performance of detecting malicious secondary users. The entire collaborative spectrum sensing process does not increase the computational and network load of spectrum sensing, nor does it incur additional performance overhead.

[0076] Furthermore, this application also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the spectrum sensing method for preventing malicious user attacks as described in any of the above embodiments.

[0077] Furthermore, this application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the spectrum sensing method for preventing malicious user attacks as described in any of the above embodiments.

[0078] See Figure 4 , Figure 4 This is a structural block diagram of a spectrum sensing device 20 for preventing malicious user attacks, provided in an embodiment of this application. The spectrum sensing device 20 includes a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21. When the processor 21 executes the computer program, it implements the steps described in the above-described embodiments of the spectrum sensing method for preventing malicious user attacks. Alternatively, when the processor 21 executes the computer program, it implements the functions of each module / unit in the above-described device embodiments.

[0079] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the spectrum sensing device 20 for preventing malicious user attacks.

[0080] The spectrum sensing device 20 for preventing malicious user attacks may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will understand that the schematic diagram is merely an example of the spectrum sensing device 20 for preventing malicious user attacks and does not constitute a limitation on the spectrum sensing device 20. It may include more or fewer components than illustrated, or combine certain components, or use different components. For example, the spectrum sensing device 20 for preventing malicious user attacks may also include input / output devices, network access devices, buses, etc.

[0081] The processor 21 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 21 is the control center of the spectrum sensing device 20 for preventing malicious user attacks, connecting all parts of the spectrum sensing device 20 through various interfaces and lines.

[0082] The processor 21 can be any one of a CPU (Central Processing Unit), GPU (Graphics Processing Unit), TPU (Tensor Processing Unit), NPU (Neural Network Processing Unit), DPU (Deep Learning Processing Unit), APU (Accelerated Processing Unit), and GPGPU (General-Purpose Computing on Graphics Processing Unit). The processor 21 is the control center of the spectrum sensing device 20 for preventing malicious user attacks, and connects various parts of the electronic device via various interfaces and lines.

[0083] The memory 22 can be used to store the computer programs and / or modules. The processor 21 implements various functions of the spectrum sensing device 20 for preventing malicious user attacks by running or executing the computer programs and / or modules stored in the memory 22 and calling the data stored in the memory 22. The memory 22 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store relevant data, etc. In addition, the memory 22 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0084] The integrated modules / units of the spectrum sensing device 20 for preventing malicious user attacks, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by the processor 21, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0085] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided in this application, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0086] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A spectrum sensing method for preventing malicious user attacks, characterized in that, include: A graph structure is constructed using secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges; The node structure data of the secondary user to be identified is extracted from the graph structure, and the node structure data is reconstructed using a trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified. The graph neural network is trained by a sample graph structure constructed with sample secondary users as nodes. Its training objective is to learn node embedding representations and edge embedding representations, which are used to identify malicious secondary users. Based on the difference between the node structure data and the reconstructed node structure data, the reconstruction error of the sub-user to be identified is obtained; Based on the reconstruction error, identify malicious secondary users from the users to be identified; Collaborative spectrum sensing is conducted using non-malicious secondary users.

2. The spectrum sensing method for preventing malicious user attacks as described in claim 1, characterized in that, The step of identifying malicious secondary users from the users to be identified based on the reconstruction error includes: Using the reconstruction error, the secondary users to be identified that meet the preset reconstruction error conditions are identified as candidate malicious secondary users; The global spectrum sensing result is obtained based on the local spectrum sensing result of the sub-user to be identified and the local spectrum sensing results of all sub-users participating in the same spectrum sensing task as the sub-user to be identified. The perception error of the candidate malicious secondary user is obtained based on the difference between the local spectrum perception result and the global spectrum perception result. Based on the perceived error, malicious secondary users are identified from the candidate malicious secondary users.

3. The spectrum sensing method for preventing malicious user attacks as described in claim 2, characterized in that, The step of identifying malicious secondary users from candidate malicious secondary users based on the perceived error includes: Obtain the similarity between the perceived error and the reconstruction error; When the similarity is less than a preset similarity threshold, obtain the distribution of neighboring candidate malicious secondary users of the candidate malicious secondary user corresponding to the candidate malicious secondary user whose similarity is less than the preset similarity threshold; Candidate malicious secondary users whose distribution satisfies the preset distribution are identified as malicious secondary users.

4. The spectrum sensing method for preventing malicious user attacks as described in claim 2, characterized in that, The step of using the reconstruction error to identify secondary users who meet the preset reconstruction error conditions as candidate malicious secondary users includes: Obtain the normalized value of the reconstruction error; When the normalized value of the reconstruction error is greater than the preset reconstruction error threshold, the user to be identified corresponding to the error threshold is identified as a candidate malicious secondary user.

5. The spectrum sensing method for preventing malicious user attacks as described in claim 3, characterized in that, The step of obtaining the similarity between the perceptual error and the reconstruction error includes: Obtain the normalized value of the perception error; The similarity is obtained by using the normalized value of the reconstruction error and the normalized value of the perception error.

6. The spectrum sensing method for preventing malicious user attacks as described in claim 1, characterized in that, The step of extracting node structure data of the secondary user to be identified from the graph structure, and reconstructing the node structure data using a trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified includes: Extract the node structure data of the sub-user to be identified from the graph structure; The attention score between the user to be identified and its neighboring users is calculated using a trained graph neural network through an attention mechanism. The attention scores are normalized to obtain the attention weights; Using the attention weights, the embedded representation of the node structure data is obtained; The embedded representation of the node structure data is reconstructed by the encoder to obtain the reconstructed node structure data.

7. A spectrum sensing device for preventing malicious user attacks, characterized in that, include: The module is used to construct a graph structure with secondary users participating in the spectrum sensing task as nodes and collaborative sensing relationships as edges. The reconstruction module is used to extract the node structure data of the secondary user to be identified from the graph structure, and reconstruct the node structure data using a trained graph neural network to obtain the reconstructed node structure data of the secondary user to be identified; wherein, the graph neural network is trained by a sample graph structure constructed with sample secondary users as nodes, and its training objective is to learn node embedding representations and edge embedding representations, which are used to identify malicious secondary users. The calculation module is used to obtain the reconstruction error of the sub-user to be identified based on the difference between the node structure data and the reconstructed node structure data. The identification module is used to identify malicious secondary users from the users to be identified based on the reconstruction error. The perception module is used to conduct collaborative spectrum sensing using non-malicious secondary users.

8. A spectrum sensing device for preventing malicious user attacks, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the spectrum sensing method for preventing malicious user attacks as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program; wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the spectrum sensing method for preventing malicious user attacks as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes a computer program / instruction that, when executed by a processor, implements the spectrum sensing method for preventing malicious user attacks as described in any one of claims 1 to 6.

Images