Wireless heterogeneous communication monitoring management system based on multi-source data

Abstract

The invention relates to the technical field of wireless communication, in particular to a wireless heterogeneous communication monitoring management system based on multi-source data. The method comprises the following steps: firstly, carrying out association modeling on a time sequence and spatial distribution of heterogeneous data by adopting a space-time diagram convolutional network, extracting network topology and equipment attribute features in combination with a knowledge graph, generating a unified feature vector, and realizing accurate network state perception; secondly, the network state is comprehensively represented by using a uniform feature vector, and an optimal spectrum allocation strategy is autonomously learned through a DQN algorithm, so that the network energy consumption is effectively reduced, the time delay is reduced, and the spectrum utilization rate is improved; finally, the spectrum allocation instruction is subjected to fully homomorphic encryption through network topology hash and equipment hash, and the target equipment can decrypt and execute only after the key is successfully matched, so that the instruction is prevented from being tampered or leaked in the transmission process, and the communication security is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communications, and in particular to a wireless heterogeneous communication monitoring and management system based on multi-source data. Background Art

[0002] With the coexistence of diverse wireless networks, including 5G, Wi-Fi, and satellite communications, heterogeneous wireless networks (HWNs) are becoming mainstream. These networks integrate diverse access technologies to improve overall performance and service quality. However, due to their heterogeneity and dynamic nature, efficient monitoring and management of these networks to improve spectrum utilization, reduce energy consumption, and ensure communication security have become a research hotspot.

[0003] Existing technologies have numerous shortcomings in managing heterogeneous wireless communication networks. Traditional methods typically rely on static or semi-dynamic strategies that are unable to adapt to rapidly changing network environments, resulting in inefficient spectrum resource utilization. Furthermore, the lack of uniformity among heterogeneous devices and protocols leads to information silos, hindering effective data fusion. Inadequate intelligent decision-making support capabilities prevent existing systems from fully exploiting spatiotemporal data features and providing precise optimization solutions. Finally, while existing encryption technologies ensure data security, they are insufficient for protecting sensitive data, such as spectrum allocation instructions, leaving them vulnerable to tampering and leakage during transmission.

[0004] Therefore, a wireless heterogeneous communication monitoring and management system based on multi-source data is proposed. Summary of the Invention

[0005] This invention provides a wireless heterogeneous communication monitoring and management system based on multi-source data. It extracts the spatiotemporal features of heterogeneous data using a spatiotemporal graph convolutional network. It then integrates network topology and device attributes with a knowledge graph to generate a unified feature vector. Based on this, the DQN algorithm is used to optimize spectrum allocation strategies, reducing network energy consumption, latency, and improving spectrum utilization. Fully homomorphic encryption is then performed using network topology hashing and device hashing to ensure secure transmission of spectrum allocation instructions.

[0006] To achieve the above object, the present invention provides the following technical solutions: Wireless heterogeneous communication monitoring and management system based on multi-source data, including: Data acquisition module, used to collect heterogeneous data, network topology and device attributes of different communication devices through multi-protocol interfaces; A data fusion module is configured to perform association modeling on the time series and spatial distribution in the heterogeneous data through a spatiotemporal graph convolutional network to obtain spatiotemporal fusion features; integrate the network topology and the device attributes through a knowledge graph to obtain an embedding vector; and integrate the spatiotemporal fusion features and the embedding vector to generate a unified feature vector; The intelligent analysis module uses a unified feature vector as input and optimizes network energy consumption, latency, and spectrum utilization to generate the optimal spectrum allocation strategy for wireless heterogeneous communication networks using the DQN algorithm. A policy encryption module is used to generate a spectrum allocation instruction according to the optimal spectrum allocation strategy, perform fully homomorphic encryption on the spectrum allocation instruction through network topology hash and device hash, and issue the encrypted instruction to the target network device by the SDN controller; The policy execution module is configured to decrypt the encrypted instruction and execute spectrum allocation after the target network device receives the encrypted instruction.

[0007] Furthermore, the spatiotemporal graph convolutional network includes: A data processing unit, configured to pre-process the heterogeneous data; A time feature extraction unit, configured to extract time series features of the heterogeneous data using LSTM; A spatial feature extraction unit is used to use the Kriging interpolation method to complete the coverage space of the heterogeneous data, and use the GCN network to extract the spatial topological features of the completed heterogeneous data; The feature fusion unit is used to fuse the time series features and spatial topology features using an attention mechanism to generate spatiotemporal fusion features.

[0008] Furthermore, integrating the network topology with the device attributes includes: Parse network topology data and device attribute data, extract entity information and relationship information, build knowledge graph triples and store them in the Neo4j graph database; TransE is used to calculate the embedding of the entity information and the relationship information to generate an embedding vector.

[0009] Furthermore, the calculation formula for generating a unified eigenvector is: ; in, represents the unified eigenvector, represents the normalized exponential activation function, represents the spatiotemporal fusion weight, represents the spatiotemporal fusion feature, represents the graph embedding weight, represents the embedding vector.

[0010] Furthermore, the step of generating the optimal spectrum allocation strategy includes: S501: Initialize the Q network and experience pool, define a unified feature vector as the state space of the network, and define the spectrum allocation scheme as the action space; S502: According to the current status, The greedy strategy selects a spectrum allocation action and executes it. The environment then feeds back the next state and the corresponding immediate reward based on the result of the action. S503: storing the current state, spectrum allocation action, immediate reward, and next state into the experience pool; S504: Randomly extract a batch of data from the experience pool, calculate the target Q value, and update the parameters of the Q network to minimize the Q value estimation error; S505: Iteratively execute steps S502-S504, and stop the iteration when the Q-value estimation error is less than the error threshold, to generate the optimal spectrum allocation strategy.

[0011] Furthermore, performing fully homomorphic encryption on the spectrum allocation instruction includes: Calculate the topology hash based on the network topology and the device hash based on the device UID; Combining the topology hash and the device hash as a key seed for a homomorphic encryption algorithm to generate an encryption key; The spectrum allocation instruction is encrypted using the CKKS homomorphic encryption algorithm to generate an encrypted instruction.

[0012] Furthermore, decrypting the encrypted instruction includes: The target network device retrieves its own device UID and network topology, calculates the topology hash and device hash, and generates the corresponding key seed; Construct the decryption key using the generated key seed and initialize the CKKS decryption environment; The target device receives the encrypted instruction and attempts to decrypt; if the key matches, the decryption succeeds, otherwise the decryption fails.

[0013] The beneficial effects of the present invention are: 1. The spatiotemporal graph convolutional network effectively extracts the temporal and spatial features of heterogeneous data, enabling more accurate association modeling and improving data integrity and expressiveness. Furthermore, it integrates network topology and device attributes with the knowledge graph to generate high-dimensional embedding vectors, enhancing semantic connections between devices. Finally, the spatiotemporal features are integrated with the embedding vectors to construct a unified feature vector, providing more comprehensive and accurate data support for subsequent spectrum allocation optimization, thereby improving network performance and resource utilization efficiency.

[0014] 2. Utilizing a unified feature vector to comprehensively characterize network status, the DQN algorithm autonomously learns the optimal spectrum allocation strategy, effectively reducing network energy consumption, latency, and spectrum utilization. Compared to traditional methods, this strategy can adapt to complex wireless heterogeneous environments, dynamically optimize resource allocation, improve network performance and communication efficiency, and achieve intelligent and efficient spectrum management.

[0015] 3. Spectrum allocation instructions are fully homomorphically encrypted using network topology hashing and device hashing, ensuring they are not tampered with or leaked during transmission, thereby enhancing communication security. The SDN controller centrally manages the issuance of encrypted instructions, making spectrum allocation more accurate and efficient. The target device can decrypt and execute only after a successful key match, enhancing access control and data integrity, ensuring the secure and efficient allocation of spectrum resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 This is a structural diagram of the wireless heterogeneous communication monitoring and management system based on multi-source data provided by the present invention; Figure 2 This is a schematic diagram of the spatiotemporal graph convolutional network structure provided by the present invention; Figure 3 This is a flow chart of the wireless heterogeneous communication monitoring and management method based on multi-source data provided by the present invention. DETAILED DESCRIPTION

[0017] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0018] Example 1 Wireless heterogeneous communication monitoring and management system based on multi-source data, such as Figure 1 As shown, it includes data acquisition module, data fusion module, data fusion module, intelligent analysis module, policy encryption module and policy execution module.

[0019] Data acquisition module, used to collect heterogeneous data, network topology and device attributes of different communication devices through multi-protocol interfaces; Specifically, heterogeneous data includes cellular network data (5G, 4G, LTE) and non-cellular network data (Wi-Fi, LoRa, Bluetooth, ZigBee). Data items include performance indicators such as frequency band, load status, signal strength, interference status, access point coverage, and signal strength distribution; network topology includes information such as device connection relationship, topological structure, and network node weight; device attributes include device model, protocol support, computing power, and other information.

[0020] A data fusion module is configured to perform association modeling on the time series and spatial distribution in the heterogeneous data through a spatiotemporal graph convolutional network to obtain spatiotemporal fusion features; integrate the network topology and the device attributes through a knowledge graph to obtain an embedding vector; and integrate the spatiotemporal fusion features and the embedding vector to generate a unified feature vector; Furthermore, the spatiotemporal graph convolutional network is as follows Figure 2 Shown, including: A data processing unit, configured to pre-process the heterogeneous data; A time feature extraction unit, configured to extract time series features of the heterogeneous data using LSTM; A spatial feature extraction unit is used to use the Kriging interpolation method to complete the coverage space of the heterogeneous data, and use the GCN network to extract the spatial topological features of the completed heterogeneous data; The feature fusion unit is used to fuse the time series features and spatial topology features using an attention mechanism to generate spatiotemporal fusion features.

[0021] The use of a spatiotemporal graph convolutional network effectively captures the temporal and spatial distribution characteristics of heterogeneous data. By combining LSTM to extract temporal features with GCN to extract spatial topological features, it not only bridges the gap between temporal and spatial data but also complements spatial data through Kriging interpolation, ensuring data integrity and accuracy. Furthermore, by fusing spatiotemporal features using an attention mechanism, the network can accurately extract spatiotemporal fusion features, improving the network's ability to understand complex dynamic environments and thereby enhancing the optimization of spectrum allocation strategies in heterogeneous wireless communication networks.

[0022] Furthermore, integrating the network topology with the device attributes includes: Parse network topology data and device attribute data, extract entity information and relationship information, build knowledge graph triples and store them in the Neo4j graph database; TransE is used to calculate the embedding of the entity information and the relationship information to generate an embedding vector.

[0023] Specifically, the ETL process is used to clean and format the raw data and store it in structured data. Data parsing tools (such as Python's pandas library) are used to convert raw log, JSON, XML and other format data into standardized formats. Stanford NER is used for entity information extraction, Open NRE is used for relationship information extraction, and rdflib is used to generate an RDF knowledge graph. Triples in the format of (entity, relationship, entity / attribute) are stored in the Neo4j graph database. TransE exports the knowledge graph triples from Neo4j and uses OpenKE or PyTorch Geometric for training to obtain low-dimensional embedding vectors.

[0024] By parsing network topology and device attribute data, extracting entity and relationship information, and constructing knowledge graph triples that are stored in the Neo4j graph database, key network information can be efficiently organized and managed. Using the TransE algorithm, entities and relationships are mapped into low-dimensional embedding vectors, improving the operability and analyzability of the data and providing accurate and reliable feature representation for subsequent network optimization and decision-making.

[0025] Furthermore, the calculation formula for generating a unified eigenvector is: ; in, represents the unified eigenvector, represents the normalized exponential activation function, represents the spatiotemporal fusion weight, represents the spatiotemporal fusion feature, represents the graph embedding weight, represents the embedding vector.

[0026] Through the above formula, the spatiotemporal fusion features and embedding vectors are weighted and integrated into a unified feature vector, and softmax is used to enhance important features and suppress unimportant features, thereby improving the representation accuracy of the network state and enhancing the accuracy and effect of the optimization strategy.

[0027] The intelligent analysis module uses a unified feature vector as input and optimizes network energy consumption, latency, and spectrum utilization to generate the optimal spectrum allocation strategy for wireless heterogeneous communication networks using the DQN algorithm. Furthermore, the step of generating the optimal spectrum allocation strategy includes: S501: Initialize the Q network and experience pool, define a unified feature vector as the state space of the network, and define the spectrum allocation scheme as the action space; S502: According to the current status, The greedy strategy selects a spectrum allocation action and executes it. The environment then feeds back the next state and the corresponding immediate reward based on the result of the action. S503: storing the current state, spectrum allocation action, immediate reward, and next state into the experience pool; S504: Randomly extract a batch of data from the experience pool, calculate the target Q value, and update the parameters of the Q network to minimize the Q value estimation error; S505: Iteratively execute steps S502-S504, and stop the iteration when the Q-value estimation error is less than the error threshold, to generate the optimal spectrum allocation strategy.

[0028] Specifically, define the unified eigenvector It is the state space of the network, reflects the communication state of the wireless heterogeneous communication network, and defines the spectrum allocation scheme. The action space is selected, that is, the matching of the channel and the device is selected according to the spectrum allocation scheme, the Q network is initialized, including the main Q network and the target Q network, and the experience pool is initialized to store the state transition data; The probability of randomly selecting an action, otherwise the action with the highest Q value is selected, and the spectrum allocation action is performed in the wireless heterogeneous communication network environment, and the environment is obtained to return the next state and instant rewards , the current state, the selected action, the reward and the next state ( ) is stored in the experience pool; a group of ( ) to train the Q network and calculate the target Q value. The calculation formula is: ; in, represents the target Q value, Indicates immediate reward, represents the discount factor, Obtained by the target Q network, indicating the next state The maximum Q value of all possible actions is obtained, and the parameters of the Q network are updated to minimize the Q value estimation error. The calculation formula of the estimation error is: ; in, represents the loss function, Indicates the number of samples drawn from the experience pool during one training session, represents the current Q value, which is obtained by the main Q network; iteratively execute steps S502-S504, stop the iteration when the Q value estimation error is less than the error threshold, and generate the optimal spectrum allocation strategy. In a feasible implementation manner, the error threshold is 0.012.

[0029] The DQN algorithm uses a unified feature vector as input, combining network energy consumption, latency, and spectrum utilization as optimization targets to gradually generate an optimal spectrum allocation strategy. Through Q network training and effective utilization of the experience pool, the spectrum allocation plan is intelligently adjusted to optimize network performance. Ultimately, by minimizing Q-value estimation errors, the spectrum allocation strategy is gradually optimized to ensure reliable resource management and efficient network applications.

[0030] A policy encryption module is used to generate a spectrum allocation instruction according to the optimal spectrum allocation strategy, perform fully homomorphic encryption on the spectrum allocation instruction through network topology hash and device hash, and issue the encrypted instruction to the target network device by the SDN controller; Furthermore, performing fully homomorphic encryption on the spectrum allocation instruction includes: Calculate the topology hash based on the network topology and the device hash based on the device UID; Combining the topology hash and the device hash as a key seed for a homomorphic encryption algorithm to generate an encryption key; The spectrum allocation instruction is encrypted using the CKKS homomorphic encryption algorithm to generate an encrypted instruction.

[0031] By combining the network topology hash and device hash to generate encryption keys, the spectrum allocation instructions are encrypted using the CKKS homomorphic encryption algorithm to ensure the secure transmission of the spectrum allocation strategy; the encrypted instructions are sent to the target network device through the SDN controller to protect data privacy and security.

[0032] The policy execution module is configured to decrypt the encrypted instruction and execute spectrum allocation after the target network device receives the encrypted instruction.

[0033] Furthermore, decrypting the encrypted instruction includes: The target network device retrieves its own device UID and network topology, calculates the topology hash and device hash, and generates the corresponding key seed; Construct the decryption key using the generated key seed and initialize the CKKS decryption environment; The target device receives the encrypted instruction and attempts to decrypt; if the key matches, the decryption succeeds, otherwise the decryption fails.

[0034] By calculating a hash based on the device UID and network topology to generate a key seed, and using the seed to construct a decryption key, it ensures that only legitimate devices can decrypt spectrum allocation instructions, thereby ensuring the security and accurate execution of the instructions.

[0035] Example 2 A communication equipment manufacturing company upgraded one of its wireless heterogeneous communication monitoring and management systems. The upgraded system was equipped with the wireless heterogeneous communication monitoring and management method based on multi-source data proposed by the present invention. Figure 3 Shown, including: Collect heterogeneous data, network topology, and device attributes from different communication devices through multi-protocol interfaces; The time series and spatial distribution in the heterogeneous data are associated and modeled using a spatiotemporal graph convolutional network to obtain spatiotemporal fusion features; the network topology and the device attributes are integrated using a knowledge graph to obtain an embedding vector; and the spatiotemporal fusion features and the embedding vector are integrated to generate a unified feature vector. Taking the unified feature vector as input and the network energy consumption, delay and spectrum utilization as optimization targets, the optimal spectrum allocation strategy for wireless heterogeneous communication networks is generated through the DQN algorithm. Generate a spectrum allocation instruction according to the optimal spectrum allocation strategy, perform fully homomorphic encryption on the spectrum allocation instruction through network topology hash and device hash, and send the encrypted instruction to the target network device by the SDN controller; After receiving the encrypted instruction, the target network device decrypts the encrypted instruction and performs spectrum allocation.

[0036] Furthermore, the spatiotemporal graph convolutional network includes: A data processing unit, configured to pre-process the heterogeneous data; A time feature extraction unit, configured to extract time series features of the heterogeneous data using LSTM; A spatial feature extraction unit is used to use the Kriging interpolation method to complete the coverage space of the heterogeneous data, and use the GCN network to extract the spatial topological features of the completed heterogeneous data; The feature fusion unit is used to fuse the time series features and spatial topology features using an attention mechanism to generate spatiotemporal fusion features.

[0037] Furthermore, integrating the network topology with the device attributes includes: Parse network topology data and device attribute data, extract entity information and relationship information, build knowledge graph triples and store them in the Neo4j graph database; TransE is used to calculate the embedding of the entity information and the relationship information to generate an embedding vector.

[0038] Furthermore, the calculation formula for generating a unified eigenvector is: ; in, represents the unified eigenvector, represents the normalized exponential activation function, represents the spatiotemporal fusion weight, represents the spatiotemporal fusion feature, represents the graph embedding weight, represents the embedding vector.

[0039] Furthermore, the step of generating the optimal spectrum allocation strategy includes: S501: Initialize the Q network and experience pool, define a unified feature vector as the state space of the network, and define the spectrum allocation scheme as the action space; S502: According to the current status, The greedy strategy selects a spectrum allocation action and executes it. The environment then feeds back the next state and the corresponding immediate reward based on the result of the action. S503: storing the current state, spectrum allocation action, immediate reward, and next state into the experience pool; S504: Randomly extract a batch of data from the experience pool, calculate the target Q value, and update the parameters of the Q network to minimize the Q value estimation error; S505: Iteratively execute steps S502-S504, and stop the iteration when the Q-value estimation error is less than the error threshold, to generate the optimal spectrum allocation strategy.

[0040] Furthermore, performing fully homomorphic encryption on the spectrum allocation instruction includes: Calculate the topology hash based on the network topology and the device hash based on the device UID; Combining the topology hash and the device hash as a key seed for a homomorphic encryption algorithm to generate an encryption key; The spectrum allocation instruction is encrypted using the CKKS homomorphic encryption algorithm to generate an encrypted instruction.

[0041] Furthermore, decrypting the encrypted instruction includes: The target network device retrieves its own device UID and network topology, calculates the topology hash and device hash, and generates the corresponding key seed; Construct the decryption key using the generated key seed and initialize the CKKS decryption environment; The target device receives the encrypted instruction and attempts to decrypt; if the key matches, the decryption succeeds, otherwise the decryption fails.

[0042] Table 1 Performance comparison of the system before and after the update

[0043] Table 1 shows a comparison of system performance before and after the update. The data in this table shows that the current system has significantly improved across multiple key performance indicators compared to the original system. First, network power consumption has been reduced from 50W to 40W, demonstrating that the current system is more efficient in spectrum allocation and resource management, thereby reducing energy consumption. Second, latency has been reduced from 120ms to 85ms, indicating that the current system can respond to network requests more quickly, improving real-time performance and user experience. Spectrum utilization has increased from 65% to 92%, demonstrating that the system is able to more fully utilize spectrum resources, reduce resource waste, and improve overall network performance. Finally, the packet loss rate has been reduced from 4% to 1%, indicating that the current system is more stable when handling network loads, reducing data loss and improving network reliability. Overall, the current system has demonstrated significant improvements in energy efficiency, latency, spectrum utilization, and network stability, demonstrating its clear superiority.

[0044] By combining a spatiotemporal graph convolutional network with a knowledge graph, this method generates a unified feature vector and uses the DQN algorithm to optimize spectrum allocation strategies, thereby improving spectrum utilization while reducing network energy consumption and latency. Experimental results show that the upgraded system outperforms the original system in terms of network energy consumption, latency, and spectrum utilization, reducing network energy consumption by 20%, latency by 29%, and spectrum utilization by 27%. The data packet loss rate also significantly decreases, demonstrating the effectiveness and superiority of the optimization strategy.

[0045] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. Wireless heterogeneous communication monitoring and management system based on multi-source data, characterized by: include: Data acquisition module, used to collect heterogeneous data, network topology and device attributes of different communication devices through multi-protocol interfaces; A data fusion module is used to perform association modeling on the time series and spatial distribution in the heterogeneous data through a spatiotemporal graph convolutional network to obtain spatiotemporal fusion features; Integrate the network topology and the device attributes through a knowledge graph to obtain an embedding vector; integrate the spatiotemporal fusion features and the embedding vector to generate a unified feature vector; The intelligent analysis module uses a unified feature vector as input and optimizes network energy consumption, latency, and spectrum utilization to generate the optimal spectrum allocation strategy for wireless heterogeneous communication networks using the DQN algorithm. A policy encryption module is used to generate a spectrum allocation instruction according to the optimal spectrum allocation strategy, perform fully homomorphic encryption on the spectrum allocation instruction through network topology hash and device hash, and issue the encrypted instruction to the target network device by the SDN controller; The policy execution module is configured to decrypt the encrypted instruction and execute spectrum allocation after the target network device receives the encrypted instruction.

2. The wireless heterogeneous communication monitoring and management system based on multi-source data according to claim 1 is characterized in that: The spatiotemporal graph convolutional network includes: A data processing unit, configured to pre-process the heterogeneous data; A time feature extraction unit, configured to extract time series features of the heterogeneous data using LSTM; A spatial feature extraction unit is used to use the Kriging interpolation method to complete the coverage space of the heterogeneous data, and use the GCN network to extract the spatial topological features of the completed heterogeneous data; The feature fusion unit is used to fuse the time series features and spatial topology features using an attention mechanism to generate spatiotemporal fusion features.

3. The wireless heterogeneous communication monitoring and management system based on multi-source data according to claim 1 is characterized in that: Integrating the network topology with the device attributes includes: Parse network topology data and device attribute data, extract entity information and relationship information, build knowledge graph triples and store them in the Neo4j graph database; TransE is used to calculate the embedding of the entity information and the relationship information to generate an embedding vector.

4. The wireless heterogeneous communication monitoring and management system based on multi-source data according to claim 1, characterized in that: The calculation formula for generating a unified eigenvector is: ； in, represents the unified eigenvector, represents the normalized exponential activation function, represents the spatiotemporal fusion weight, represents the spatiotemporal fusion feature, represents the graph embedding weight, represents the embedding vector.

5. The wireless heterogeneous communication monitoring and management system based on multi-source data according to claim 1 is characterized in that: The step of generating the optimal spectrum allocation strategy includes: S501: Initialize the Q network and experience pool, define a unified feature vector as the state space of the network, and define the spectrum allocation scheme as the action space; S502: According to the current status, The greedy strategy selects a spectrum allocation action and executes it. The environment then feeds back the next state and the corresponding immediate reward based on the result of the action. S503: storing the current state, spectrum allocation action, immediate reward, and next state into the experience pool; S504: Randomly extract a batch of data from the experience pool, calculate the target Q value, and update the parameters of the Q network to minimize the Q value estimation error; S505: Iteratively execute steps S502-S504, and stop the iteration when the Q-value estimation error is less than the error threshold, to generate the optimal spectrum allocation strategy.

6. The wireless heterogeneous communication monitoring and management system based on multi-source data according to claim 1, characterized in that: Performing fully homomorphic encryption on the spectrum allocation instruction includes: Calculate the topology hash based on the network topology and the device hash based on the device UID; Combining the topology hash and the device hash as a key seed for a homomorphic encryption algorithm to generate an encryption key; The spectrum allocation instruction is encrypted using the CKKS homomorphic encryption algorithm to generate an encrypted instruction.

7. The wireless heterogeneous communication monitoring and management system based on multi-source data according to claim 1, characterized in that: Decrypting the encrypted instruction includes: The target network device retrieves its own device UID and network topology, calculates the topology hash and device hash, and generates the corresponding key seed; Construct the decryption key using the generated key seed and initialize the CKKS decryption environment; The target device receives the encrypted instruction and attempts to decrypt; if the key matches, the decryption succeeds, otherwise the decryption fails.

Images