Dynamic bandwidth allocation and secure transmission method for multi-chain network

Abstract

The invention discloses a dynamic bandwidth allocation and secure transmission method oriented to a multi-chain network, belongs to the technical field of network communication, and aims to solve the problems of high delay, low security and response misalignment of the multi-chain network caused by static bandwidth allocation, lagging topology perception and splitting of a security mechanism in the prior art. The method comprises the following steps: collecting data of each chain and an inter-chain channel in a multi-chain network as multi-dimensional heterogeneous data; performing feature extraction on each dimension of heterogeneous data in the multi-dimensional heterogeneous data, and performing entanglement fusion on the extracted features to obtain fused features; and based on the fusion features, obtaining a bandwidth allocation matrix and a path planning tensor through the trained dynamic bandwidth allocation and secure transmission network. According to the invention, the problems of high delay, low security and response misalignment of a multi-chain network are solved.

Description

Technical Field

[0001] This invention belongs to the field of network communication technology, specifically relating to a dynamic bandwidth allocation and secure transmission method, system, and electronic device for multi-chain networks. Background Technology

[0002] The technical challenges of multi-chain networks are gradually revealing their deep-seated contradictions. From a resource allocation perspective, the static allocation strategy in traditional multi-chain networks presents an irreconcilable contradiction with the ever-changing and highly dynamic business demands. Currently prevalent weighted polling or fixed bandwidth reservation mechanisms are proving inadequate in the face of sudden and non-linear traffic increases. For example, in scenarios like DeFi liquidation, where timeliness is extremely critical, the system needs to respond within milliseconds. However, the detection latency caused by periodic polling can be several seconds, directly leading to numerous request timeouts and potentially triggering a chain reaction of liquidation failures. Worse still, fixed priority mechanisms cannot accurately distinguish the urgency of security patch transmissions from regular log synchronization, causing critical updates to be blocked by low-priority tasks, significantly extending the window of vulnerability exposure. Furthermore, the chain-like dependencies inherent in cross-chain communication cause localized congestion to spread rapidly, resulting in resource fragmentation with idle upstream resources and scarce downstream resources, further reducing overall network efficiency.

[0003] This imbalance at the resource level, coupled with the lag in topology awareness, exacerbates the vulnerability of multi-chain systems. Most existing graph neural networks (GNNs) are built on static topology snapshots, making it difficult to capture dynamic behaviors such as frequent node online / offline cycles or malicious drift. Attackers exploit the delay in topology updates to launch "topology drift attacks," first posing as a high-bandwidth node to attract critical traffic, then suddenly going offline, causing routing oscillations and asset transfers. Meanwhile, whenever the network structure changes, traditional GNNs require several hours to retrain. During this period, routing decisions deviate significantly from reality, leading to problems such as insufficient bandwidth on new chains and continued access to old chains. Particularly in long-distance cross-chain paths, due to the limitations of local neighborhoods in GNNs, information transmission is significantly attenuated, and changes in the load of remote chains cannot be detected in a timely manner. For critical services like medical image retrieval, this results in delays several times greater, severely impacting service quality.

[0004] A deeper challenge lies in the disconnect between security mechanisms and efficiency optimization. Current security protocols, such as TLS, operate independently of resource scheduling, sacrificing performance and creating security vulnerabilities. With the looming threat of quantum computing, long-term valid session keys face the risk of being backtracked and cracked, while frequent key changes further increase handshake overhead. Advanced verification technologies like zero-knowledge proofs, while improving security, introduce additional latency due to their independent execution, which accumulates significantly in high-concurrency scenarios, directly increasing the clearing failure rate. Simultaneously, the encryption process itself consumes substantial bandwidth resources—TLS handshakes occupy one-fifth of the link capacity, and zero-knowledge proofs nearly double the data volume, drastically reducing the network's effective transmission capacity. These three bottlenecks are not isolated but form a vicious cycle of "rigidity → congestion → topology change → delayed perception → security pressure → resource crowding," severely hindering the reliable application of multi-chain systems in critical fields such as finance, healthcare, government affairs, and 5G equipment supply. Only by achieving a three-pronged innovation—dynamic resource scheduling, real-time topology awareness, and secure lightweight integration—can the current technological impasse be broken.

[0005] Based on this, the present invention proposes a dynamic bandwidth allocation and secure transmission method for multi-chain networks, which achieves millisecond-level dynamic optimization through a three-in-one architecture of chaotic-quantum collaborative sensing, dynamic graph network topology modeling, and secure transmission. Summary of the Invention

[0006] To address the aforementioned problems in the prior art, namely, to resolve the issues of high latency, low security, and inaccurate response in multi-chain networks caused by static bandwidth allocation, delayed topology awareness, and fragmented security mechanisms, the first aspect of this invention proposes a dynamic bandwidth allocation and secure transmission method for multi-chain networks, comprising: S10, Collect data from each chain and inter-chain channels in the multi-chain network as multi-dimensional heterogeneous data; the multi-dimensional heterogeneous data includes topology data, traffic data, and event data; S20, feature extraction is performed on each dimension of the heterogeneous data in the multidimensional heterogeneous data, and the extracted features are entangled and fused to obtain fused features: Based on the topology data, topology features are extracted using graph isomorphic network (GIN); based on the flow data, a singularity spectrum is obtained using multifractal analysis (MFA); based on the singularity spectrum and the flow data, self-similar features are obtained using MFA; based on the event data, the probability distribution of each type of event is calculated; based on the probability distribution, quantum states are constructed and spliced ​​together to form quantum event states. Entangling and fusing the topological features, the self-similar features, and the quantum event states through quantum gate operations yields the fused features: The quantum event state is converted into a real number vector and then tensor-producted with the topological feature to obtain the first feature; the sum of the first feature and the self-similar feature is calculated and compressed using the ReLU activation function, and the compressed feature is used as the fusion feature; S30, based on the fusion features, obtain the bandwidth allocation matrix and path planning tensor through the trained dynamic bandwidth allocation and secure transmission network; the dynamic bandwidth allocation and secure transmission network is constructed based on a deep neural network.

[0007] In some preferred embodiments, based on the topology data, topological features are extracted using Graph Isomorphic Network (GIN), the method being as follows: Based on the topology data, the degree of each node in the topology structure is obtained; the degree of two nodes and the communication time interval between the two nodes are concatenated into vectors, and then processed by a multilayer perceptron and an activation function in sequence to obtain the link survival probability between the two nodes. A mask matrix is ​​constructed based on the link survival probability between the two nodes; the Hadamard product between the static connection matrix between the two nodes and the mask matrix is ​​calculated, and then added to the event-driven adjacency matrix between the two nodes to obtain the dynamic adjacency matrix; The degree of each node is concatenated with the encoded vector of the node type information. The concatenated vector is then multiplied with the set parameter matrix and processed by the activation function to obtain the initial feature vector of each node. The initial feature vectors of each node and the dynamic adjacency matrix are input into the GNN network, and the outputs of different layers of the GNN network are fused to obtain fused features. Based on the fusion features, the average feature value of all nodes in each chain in the multi-chain network after MLP processing is calculated; the average feature value of other chains is weighted according to the attention weight between the chain to which the current node belongs and other chains, and added to the fusion feature of the current node as the enhanced feature; Historical features are obtained by performing dilated causal convolution on the historical topological data of the current node. The historical features and the enhanced features are then concatenated and combined with a learnable parameter matrix and an activation function to obtain the sensitivity weights of the current node to events. The historical features and the enhanced features are then weighted based on the sensitivity weights to obtain temporal enhanced features, which are the final extracted topological features.

[0008] In some preferred embodiments, the bandwidth allocation matrix and path planning tensor are obtained through a trained dynamic bandwidth allocation and secure transmission network. The method is as follows: The fused features are mapped to the chaotic space using a linear transformation as the initial values ​​for the Lorenz differential equation; the Lorenz differential equation is iterated and solved using the four-section Runge-Kutta method to obtain the trajectory point sequence; the state of each node at each time point in the chaotic space is sampled from the trajectory point sequence as the chaotic state C; The chaotic state C is compressed and an entangled state is constructed; based on the entangled state, the quantum mutual information between node pairs is calculated, and then the quantum correlation matrix Q is constructed; Combining the quantum correlation matrix Q, a time-varying input current is constructed through chaotic gating modulation; based on the time-varying input current, the change of the mesenchymal membrane potential is simulated through the Izhikevich neuron model, and a pulse sequence S is output. Based on the pulse sequence S, the pulse energy integral E is calculated; if the pulse energy integral exceeds the set energy threshold, a phase transition is performed on the pulse energy integral, and the phase transition result is elastically embedded and deformed to update the original topology matrix of each node, thereby obtaining the updated topology matrix. Combining the updated topology matrix, the chaotic state, the quantum correlation matrix, and the pulse sequence, the bandwidth allocation matrix and path planning tensor are output.

[0009] In some preferred embodiments, the bandwidth allocation matrix is ​​obtained by: Based on the updated topology matrix, the chaotic state, and the quantum correlation matrix, the initial bandwidth allocation result is determined; Based on the established physical constraints and the initial bandwidth allocation results, the optimal bandwidth allocation results are calculated using the projection gradient descent method. Then, the bandwidth allocation matrix is ​​obtained by dividing the bandwidth allocation into different levels.

[0010] In some preferred embodiments, the path planning tensor is obtained by: The bandwidth allocation matrix, the risk warning vector, and the quantum correlation matrix are used to obtain a path planning tensor, i.e., the optimal path or candidate path, by employing a multi-objective shortest path algorithm.

[0011] In some preferred embodiments, a multi-objective shortest path algorithm is used to obtain the path planning tensor, and the method is as follows: Based on the bandwidth allocation matrix, the risk warning vector, and the quantum correlation matrix, obtain the path available bandwidth, path failure probability, path transmission time, and the minimum value in the quantum correlation matrix corresponding to all edges of each path in the current network; perform a weighted summation of the path available bandwidth, path failure probability, path transmission time, and the minimum value in the quantum correlation matrix corresponding to all edges of each path to obtain the initial path value. Based on the degree of each node in each path, the sum of local degrees, and the sum of the degrees of all nodes in the network, the multi-scale topological entropy corresponding to each path is calculated; the contribution of the topological entropy is determined according to the ratio of each multi-scale topological entropy to the maximum value of the multi-scale topological entropy of all paths in the current network; the sum of local degrees is the sum of the degrees of all nodes in each path. Calculate the similarity between each path and the paths in the selected path set, multiply each similarity by the corresponding time decay factor and sum them up, and take the difference between 1 and the sum as the contribution of the path interaction entropy. The initial path value, the contribution of the topological entropy, and the contribution of the path interaction entropy are multiplied together to obtain the final path value for each path. The path corresponding to the maximum value of the final path value is selected as the optimal path, or the path corresponding to the maximum value of a predetermined number of paths is selected as the candidate path.

[0012] In some preferred embodiments, the topological features, the self-similar features, and the quantum event states are entangled and fused using quantum gate operations to obtain fused features. The method is as follows:

[0013] in, Indicates fusion features, Representing topological features, Indicates self-similarity features, Representing quantum event states, This represents the weight vector set during feature fusion.

[0014] In a second aspect, the present invention proposes a dynamic bandwidth allocation and secure transmission system for multi-chain networks. Based on the aforementioned dynamic bandwidth allocation and secure transmission method for multi-chain networks, the system includes: The data acquisition module is configured to collect data from each chain and inter-chain channels in the multi-chain network as multi-dimensional heterogeneous data; the multi-dimensional heterogeneous data includes topology data, traffic data, and event data. The feature extraction module is configured to extract features from each dimension of the multidimensional heterogeneous data, and then entangle and fuse the extracted features to obtain fused features. Based on the topology data, topology features are extracted using graph isomorphic network (GIN); based on the flow data, a singularity spectrum is obtained using multifractal analysis (MFA); based on the singularity spectrum and the flow data, self-similar features are obtained using MFA; based on the event data, the probability distribution of each type of event is calculated; based on the probability distribution, quantum states are constructed and spliced ​​together to form quantum event states. Entangling and fusing the topological features, the self-similar features, and the quantum event states through quantum gate operations yields the fused features: The quantum event state is converted into a real number vector and then tensor-producted with the topological feature to obtain the first feature; the sum of the first feature and the self-similar feature is calculated and compressed using the ReLU activation function, and the compressed feature is used as the fusion feature; The bandwidth allocation and secure transmission module is configured to obtain a bandwidth allocation matrix and a path planning tensor based on the fusion features and through a trained dynamic bandwidth allocation and secure transmission network; the dynamic bandwidth allocation and secure transmission network is constructed based on a deep neural network.

[0015] In a third aspect, the present invention provides an electronic device comprising: At least one processor; and A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement the above-described dynamic bandwidth allocation and secure transmission method for multi-chain networks.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer instructions for execution by a computer to implement the above-described method for dynamic bandwidth allocation and secure transmission in multi-chain networks.

[0017] The beneficial effects of this invention are: This invention solves the problems of high latency, low security, and inaccurate response in multi-chain networks.

[0018] 1) This invention overcomes the nonlinear constraint problem in traditional resource allocation by dynamically driving the chaotic field equation through quantum entanglement correlation. This mechanism quantifies cross-domain cooperative relationships into physically computable entanglement parameters, directly controls the bandwidth allocation strategy, realizes the fundamental transformation from static preset to dynamic real-time response, and significantly enhances the resource guarantee capability of high-priority tasks. 2) This invention integrates event-triggered logic with multi-scale graph neural network aggregation, and combines hierarchical fusion of local and global features to construct a closed-loop system for topology dynamic perception and feature extraction. This completely solves the response lag problem caused by retraining in traditional models, realizes real-time reinforcement of critical paths, and effectively maintains the spatiotemporal consistency of network state. 3) This invention maximizes the transmission efficiency of multi-chain networks while ensuring security and reliability, completely solves the problem of mutual exclusion between security, efficiency and cost in multi-chain networks, improves the accuracy of response and reduces latency. Attached Figure Description

[0019] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating a dynamic bandwidth allocation and secure transmission method for multi-chain networks according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0022] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0023] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0024] A method for dynamic bandwidth allocation and secure transmission in multi-chain networks according to a first embodiment of the present invention, such as... Figure 1 As shown, the method includes: S10, Collect data from each chain and inter-chain channels in the multi-chain network as multi-dimensional heterogeneous data; the multi-dimensional heterogeneous data includes topology data, traffic data, and event data; S20, feature extraction is performed on each dimension of the heterogeneous data in the multidimensional heterogeneous data, and the extracted features are entangled and fused to obtain fused features: Based on the topology data, topology features are extracted using graph isomorphic network (GIN); based on the flow data, a singularity spectrum is obtained using multifractal analysis (MFA); based on the singularity spectrum and the flow data, self-similar features are obtained using MFA; based on the event data, the probability distribution of each type of event is calculated; based on the probability distribution, quantum states are constructed and spliced ​​together to form quantum event states. Entangling and fusing the topological features, the self-similar features, and the quantum event states through quantum gate operations yields the fused features: The quantum event state is converted into a real number vector and then tensor-producted with the topological feature to obtain the first feature; the sum of the first feature and the self-similar feature is calculated and compressed using the ReLU activation function, and the compressed feature is used as the fusion feature; S30, based on the fusion features, obtain the bandwidth allocation matrix and path planning tensor through the trained dynamic bandwidth allocation and secure transmission network; the dynamic bandwidth allocation and secure transmission network is constructed based on a deep neural network.

[0025] To more clearly illustrate the dynamic bandwidth allocation and secure transmission method for multi-chain networks according to the present invention, the steps of one embodiment of the system of the present invention will be described in detail below with reference to the accompanying drawings.

[0026] S10, Collect data from each chain and inter-chain channels in the multi-chain network as multi-dimensional heterogeneous data; the multi-dimensional heterogeneous data includes topology data, traffic data, and event data; In this embodiment, data from each chain and inter-chain channels in the multi-chain network are first collected, including topology data, traffic data, and event data. The multi-chain network is the multi-chain network of the enterprise's 5G equipment supply chain. The chain network includes at least three chains: a development chain (i.e., a software algorithm development chain), a testing chain (hardware / network function testing chain), and a production chain (equipment mass production deployment chain). It may also include an operation and maintenance service chain and a hardware intelligent manufacturing chain. The topology data is the chain network physical connection matrix, which is an N×N binary matrix where N is the number of chains. Therefore, the topology data (taking 3 chains as an example) is the link relationship matrix of the development chain, testing chain, and production chain. The traffic data is a time-series traffic feature tensor T×N×F, where T represents the total number of time steps and F represents the feature dimension. It is used to record resource usage or behavior indicators that change on the chain over time, such as at least code submission frequency, deployment delay, and test case execution time. The event data is a discrete event code T×N×K on the chain, where K represents the number of event types, such as at least security vulnerability alerts, version release time, and compliance audit status.

[0027] S20, feature extraction is performed on each dimension of the heterogeneous data in the multidimensional heterogeneous data, and the extracted features are entangled and fused to obtain fused features; In this embodiment, because the heterogeneous data in each dimension of the multidimensional heterogeneous data differ significantly, it is necessary to first perform encoding unification processing before performing quantum fusion processing. Specifically, as follows: S201, based on the topology data, extract topology features using Graph Isomorphic Network (GIN); based on the flow data, obtain the singularity spectrum using Multifractal Analysis (MFA); based on the singularity spectrum and the flow data, obtain self-similar features using MFA; based on the event data, calculate the probability distribution of each type of event; based on the probability distribution, construct and concatenate quantum states as quantum event states. In this invention, topological features are extracted using graph isomorphic networks (GIN), specifically as follows: Based on the topology data, the degree of each node in the topology structure is obtained; the degree of two nodes and the communication time interval between two nodes (i.e., the time interval from the last communication between the two nodes to the present, the longer the time, the higher the probability of link failure) are concatenated into vectors, and then processed sequentially through a multilayer perceptron (existing technology, including an input layer, a hidden layer, and an output layer, the input dimension of the output layer is determined according to the input data, the hidden layer is generally 1 layer with 64 neurons, and the output layer is determined according to the output dimension) and an activation function (preferably the ReLU function) to obtain the link survival probability between the two nodes; A mask matrix is ​​constructed based on the link survival probability between the two nodes, where each element of the mask matrix represents the link survival probability between the two nodes. The Hadamard product between the static connection matrix between the two nodes and the mask matrix is ​​calculated and added to the event-driven adjacency matrix between the two nodes to obtain the dynamic adjacency matrix. In this invention, the static connection matrix between two nodes The initial network connection structure, That is, if two nodes are connected in the initial topology, it is set to 1, otherwise it is set to 0; An event-driven adjacency matrix between two nodes is a connection temporarily established by a specific trigger (such as a security threat, failure, or external notification). Specifically, when the set triggering condition is met, the edge between the affected nodes is set to 1, indicating that a new connection or path has been triggered by a security event at the current time. Otherwise, the corresponding matrix element remains 0.

[0028] The degree of each node is concatenated with the encoded vector of the node type information. The concatenated vector is then multiplied by the set parameter matrix (i.e., weight matrix) and processed by the activation function to obtain the initial feature vector of each node. The initial feature vectors of each node and the dynamic adjacency matrix are input into the GNN network (the initial feature vectors provide basic information about each node, and the dynamic adjacency matrix defines the information propagation method (i.e., which nodes are considered neighbors). In each GNN operation, the dynamic adjacency matrix selects the neighbor nodes to be aggregated and updates the features of the current node based on the features of these neighbor nodes. The GNN network consists of three parts: 1-hop local aggregation, 2-hop local aggregation, and global aggregation. The inputs of each part are the initial feature vectors of each node and the dynamic adjacency matrix. 1-hop local aggregation is preferably set to 2 GAT layers, with 4 attention heads per layer, LeakyReLU attention mechanism, and 1 shortest path hop. 2-hop local aggregation is preferably set to 2 GCN layers, with 64 filters, ReLU activation function, and 2 shortest path hops. Global aggregation uses Graph Attention Pooling (GAP) or a global attention mechanism. Finally, the outputs of the three parts are fused. The outputs of different layers of the GNN network are fused to obtain fused features. Based on the fusion features corresponding to each node, the average feature value of all nodes in each chain of the multi-chain network after MLP processing is calculated (i.e., the fusion features corresponding to each node in each chain of the multi-chain network are processed by MLP (the hidden layer of the multilayer perceptron is also set to 64 neurons), accumulated and then averaged); the average feature value of other chains is weighted according to the attention weight between the chain to which the current node belongs and other chains, and added to the fusion features of the current node as the enhanced feature; Historical features are obtained by performing dilated causal convolution on the historical topological data of the current node (preferably with 3 layers of convolution, kernel size of 3, dilation rates of 1, 2, and 4 respectively, receptive field of 13 time steps, and activation function preferably ReLU). The historical features and the enhanced features are then concatenated, and the concatenated features are combined with a learnable parameter matrix and an activation function (i.e., the concatenated features are first weighted by the learnable parameter matrix and then activated) to obtain the sensitivity weight of the current node to events. The historical features and the enhanced features are then weighted based on the sensitivity weight to obtain the temporal enhanced features, which are the final extracted topological features.

[0029] Based on the flow data, the singularity spectrum is obtained using the existing multifractal analysis method (MFA); for details on the multifractal analysis method, please refer to... or 2. [1] Kantelhardt JW , Zschiegner SA , Koscielny-Bunde E ,etal.Multifractaldetrended fluctuation analysis of nonstationary time series[J].Physica A Statistical Mechanics&Its Applications, 2002, 316(1-4).DOI:10.1016 / S0378-4371(02)01383-3. Based on the singularity spectrum and the flow data, self-similar features are obtained using the multifractal analysis method (MFA). , Represents traffic data, This is the singularity spectrum. Alternatively, in other embodiments, traffic data can be directly input into an LSTM model (e.g., selecting a 1-layer LSTM, setting the number of hidden units to 256, selecting the tanh function as the activation function, or using a conventional LSTM model first, adjusting the model parameters according to the actual situation) to extract features as self-similar features.

[0030] For each type of event, calculate the probability distribution; based on the probability distribution, construct the quantum state as shown in the following formula:

[0031] in, Representing quantum states, This represents the probability distribution of each chain node at each time step (i.e., the normalized probability distribution is calculated based on the counts of all event types at each chain node at each time step). represents the time step, and n represents the chain node.

[0032] Finally, for each The corresponding quantum states are spliced ​​into a matrix. , as a quantum event state.

[0033] S202, the topological features, the self-similar features, and the quantum event states are entangled and fused through quantum gate operations to obtain fused features: The quantum event state is converted into a real number vector, and a tensor product is calculated with the topological feature to obtain the first feature. The sum of the first feature and the self-similar feature is calculated, and the sum is compressed using the ReLU activation function. The compressed feature is used as the fusion feature, as shown in the following formula:

[0034] in, Indicates fusion features, Representing topological features, Indicates self-similarity features, Representing quantum event states, This represents the weight vector set during feature fusion.

[0035] S30, based on the fusion features, obtain the bandwidth allocation matrix and path planning tensor through the trained dynamic bandwidth allocation and secure transmission network; the dynamic bandwidth allocation and secure transmission network is constructed based on a deep neural network.

[0036] In this embodiment, based on the fusion features, the bandwidth allocation matrix and path planning tensor are obtained through the trained dynamic bandwidth allocation and secure transmission network, as follows: S301, use a linear transformation to map the fused features to the chaotic space as the initial value of the Lorenz differential equation; iterate the Lorenz differential equation and solve it using the four-section Runge-Kutta method to obtain the trajectory point sequence; sample the state of each node at each time point in the chaotic space from the trajectory point sequence as the chaotic state C; In this invention, fusion features are achieved through linear transformation. Mapped to the initial values ​​of the Lorenz differential equation ,Right now , Let T denote the learnable parameter matrix, and T denote the transpose.

[0037] S302, the chaotic state C is compressed and an entangled state is constructed; based on the entangled state, the quantum mutual information between node pairs is calculated, and then the quantum correlation matrix Q is constructed; In this invention, it is preferable to compress the chaotic state using a multilayer perceptron (the hidden layer of which is also set to 64 neurons), and then compress the resulting vector. Mapped to a set of quantum gate parameters ,in, represents the learnable parameter matrix corresponding to the compressed vector, and represents the bias term.

[0038] By combining the quantum gate parameters of all nodes, a global entanglement operation is constructed. This invention preferably uses the CNOT quantum bit gate and the CPHASE quantum bit gate to perform entanglement operation using two quantum gate parameters. After the entanglement operation, the result of the entanglement operation is multiplied by the pre-constructed initial maximally entangled state to obtain the final entangled state; wherein, the initial entangled state is preferably: .

[0039] Construct the density matrix based on the final entangled state. For the density matrix The partial trace is calculated to obtain the joint reduced density matrix. The matrix is ​​then divided into blocks according to nodes to obtain the reduced density matrix for each node. Based on each reduced density matrix, the von Neumann entropy is calculated. The sum of the von Neumann entropies corresponding to each node is then taken as the difference between the sum and the von Neumann entropy corresponding to the joint reduced density matrix, which is used as the quantum mutual information between the two nodes. The quantum mutual information between all nodes at each time step is constructed into a matrix, which serves as the quantum mutual information for that time step. t The quantum correlation metric of the network topology, namely the quantum correlation matrix.

[0040] S303, combined with the quantum correlation matrix Q, a time-varying input current is constructed through chaotic gating modulation; based on the time-varying input current, the change of the mesenchymal membrane potential is simulated through the Izhikevich neuron model, and a pulse sequence S is output; In this invention, a time-varying input current is constructed. The method is as follows:

[0041] in, The angular frequency of the input signal is represented by t, which is the event data. The time step is represented by t.

[0042] S304, Based on the pulse sequence S, calculate the pulse energy integral E; if the pulse energy integral exceeds the set energy threshold, perform a phase transition on the pulse energy integral, and perform elastic embedding deformation on the phase transition result, thereby updating the original topology matrix of each node to obtain the updated topology matrix. In this invention, a pulse sequence over a period of time is integrated to obtain the pulse energy integral E. If the pulse energy integral exceeds a set energy threshold (preferably set to 65), a phase transition is performed on the pulse energy integral, including austenitic and martensitic phase transitions. The vectors after the phase transition are concatenated and weighted, and then processed by an activation function to obtain the elastic embedding deformation result. The elastic embedding deformation result is transformed into an incremental matrix (for example, the elastic embedding deformation results of two nodes can be concatenated using a neural network, and then processed by an MLP and activated to obtain an incremental matrix), and then the original topology matrix of each node is updated.

[0043] S305, combining the updated topology matrix, the chaotic state, the quantum correlation matrix, and the pulse sequence, outputs a bandwidth allocation matrix and a path planning tensor.

[0044] In this invention, the bandwidth allocation matrix is ​​obtained as follows: Based on the updated topological matrix, chaotic state, and quantum correlation matrix, the initial bandwidth allocation is determined. Combining this with the established physical constraints, the optimal bandwidth allocation is calculated using the projective gradient descent method, and then divided according to different levels to obtain the bandwidth allocation matrix. The details are as follows: First, the chaotic states and quantum correlation matrices of the two nodes are fused using a fusion function (e.g., at least weighted average fusion and exponential weighted fusion, where weighted fusion is: , , The weighting parameters are for weighting. , This represents the chaotic state of two nodes. (This represents the quantum correlation matrix between the two nodes), and then the fused result is multiplied by the updated topology matrix to obtain the initial bandwidth allocation result.

[0045] The physical constraints set include total bandwidth conservation (i.e., the sum of the bandwidths of all links is the total bandwidth) and single-link bandwidth limits (i.e., the bandwidth of a single link can be greater or less than a certain value).

[0046] Based on the set physical constraints and the initial bandwidth allocation results, the optimal bandwidth allocation results are calculated using the projection gradient descent method. The optimal bandwidth allocation results are then divided into different bandwidth levels (e.g., at least low, medium, and high) to form the final three-dimensional matrix, which serves as the bandwidth allocation matrix.

[0047] The path planning tensor is obtained by using a multi-objective shortest path algorithm based on the bandwidth allocation matrix, risk warning vector, and quantum correlation matrix to obtain candidate paths, i.e., the path planning tensor.

[0048] In this invention, when calculating the shortest path for multiple objectives, multi-scale entropy, time decay factor, and path interaction entropy are introduced to break through the traditional single entropy model, improve path diversity, reduce traffic concentration in wind direction, and dynamically adapt to historical path selection to avoid duplicate paths. The objective function of the multi-objective shortest path algorithm is as follows:

[0049]

[0050] Where score(p) represents the path value corresponding to path p, i.e., the final path value, and H(P) represents the multi-scale topological entropy. This represents the maximum multi-scale topological entropy of all paths in the current network. Let p be the existence time of path m, expressed in time steps. Let sim(p,m) represent the path similarity. The time decay factor is δ, which represents the rate of decay. The path transmission time, path failure probability, path available bandwidth, and the corresponding values ​​for all edges on the path are given. The initial path value is obtained by weighted summation of the minimum values ​​in the quantum correlation matrix (i.e., the minimum values ​​in the quantum correlation matrix). The available bandwidth of the path is obtained through the bandwidth allocation matrix. The risk warning vector is a matrix formed by combining the monitored vectors according to their dimensions, based on the traffic, latency, and packet loss rate between network nodes. The path failure probability is obtained through the risk warning vector. For example, by obtaining the traffic, latency, and packet loss rate of each node, the path failure probability can be obtained through commonly used neural networks, which is a common existing technology and will not be detailed here. In other embodiments, other existing technologies can be used to obtain the path failure probability. d(v) represents the degree of node v. This represents the sum of local degrees, that is, the sum of the degrees of all nodes in path p. The sum of the degrees of all nodes in the network is represented by α, which is the balance coefficient between local and global entropy, preferably 0.1 in this invention; β is the adjustment coefficient, preferably 1 in this invention; and γ is the penalty intensity for controlling the path interaction entropy, preferably 0.5 in this invention. This represents the set of paths that have been selected in the current or previous selection process, i.e. It starts as an empty set, and once a path is selected, it is added to the set.

[0051] Finally, the path corresponding to the largest score(p) value is selected as the final selected path, or the paths corresponding to the largest score(p) values ​​of the previous set number are retained as candidate paths.

[0052] In summary, this invention fundamentally solves the three major technical bottlenecks of rigid resource allocation in multi-chain networks, lagging topology awareness, and fragmented security mechanisms. Verification has shown that the method of this invention reduces the average task latency of the 5G equipment supply chain by 72% and reduces security incidents by 92%, providing a revolutionary solution for multi-chain infrastructure.

[0053] To facilitate understanding of the above technical solutions, the following example uses Company A's 5G equipment supply chain: Company A's 5G equipment supply chain comprises a multi-chain network with three chains (development, testing, and production). Real-time monitoring of its topology, traffic characteristics, and security events is required to detect traffic anomalies (such as DDoS attacks) and dynamically adjust bandwidth and paths to ensure high availability of the production chain. Details are as follows: S10, the topology data is a 3×3 two-dimensional matrix, representing the link relationships between chains. 1 indicates a direct relationship between chains, and 0 indicates no connection. Specifically, the first row of the matrix represents direct relationships between the development chain and the test chain, and between the development chain and the production chain; the second row represents direct relationships between the test chain and the development chain, and between the test chain and the production chain; the third row represents direct relationships between the production chain and the test chain. Initial bandwidth allocation: all links have a default bandwidth of 200Mbps (total bandwidth 1200Mbps); traffic data: T=10 (time steps, 1 minute per step), N=3 (number of chains), F=4 (feature dimension). Each time step records four types of metrics on the chain (such as code commit frequency (times / minute), deployment latency (seconds)). Test case time (seconds), resource utilization (percentage)), for example, at time step 0, the development chain traffic data is: [100,5,200,80], the test chain traffic data is [80,10,150,70], and the production chain traffic data is [50,15,100,60]. Event data T=10 (number of time steps, 1 minute per step), N=3 (number of chains), K=3 (event type: security vulnerability, version release, compliance audit). For example, at time step 0, the development chain triggers a version release event, and the event data is [0,1,0]. The test chain triggers compliance design, and the event data is [0,0,1]. The production chain triggers a security vulnerability, and the event data is [1,0,0].

[0054] S20, obtain the degree [2,2,2] of each node, i.e., development chain degree = 2 (linking to test chain and production chain), test chain degree = 2 (linking to development chain and production chain), and production chain degree = 2 (linking to test chain); then, calculate the communication time interval (development chain links to test chain, current time is step 10, the development chain and test chain last communicated at step 8, communication time interval is 2, so M[0,1] = 1 - 2 / 10 = 0.8; development chain links to production chain, last communicated at step 9, communication time interval is 1, so M[0.2] = 0.9; test chain links to production chain, last communicated at step 7, communication time interval is 3, so M[1,2] is 0.7), construct the mask matrix. The diagonal is 1; Combining the mask matrix and the static connection matrix ( ), calculate the Hadamard product, matrix The event-driven adjacency matrix (t=0 triggers a security vulnerability) is... (Adding new links between the production chain and the development chain), the dynamic adjacency matrix is ​​as follows: The normalized dynamic adjacency matrix is Then, based on the dynamic adjacency matrix and the initial node features (i.e., degree + type encoding, type encoding: development = 001, test = 010, production 100), the initial feature vector [2, 0, 0, 1] of the development chain is obtained, which is then input into the GNN network to finally obtain the extracted topological features. Calculate the singularity spectrum for the traffic data and extract self-similar features (e.g., for 10 time steps [100,120,130,140,150,160,170,180,190,150], q∈[-5,5], calculate the singularity spectrum). =0.2 (Pay attention to small fluctuations) =0.5 (equilibrium point) =0.65, the self-similarity feature is 0.65 (when q=2); or input the traffic data into the LSTM network to obtain the self-similarity feature; The event data is normalized to obtain a probability distribution (e.g., if the time step is 0, the development chain and security vulnerabilities are 0, the version release is 1, and the compliance audit is 0, then the event count is [0,1,0], and the normalized probability distribution is [0,1,0]; the test chain is [0,0,1], and the production chain is [1,0,0]). Then, quantum state construction is performed (taking the development chain as an example). This leads to the quantum event state. Finally, quantum entanglement fusion is performed to obtain the fusion characteristics; S30, with initial values ​​[0.45, 0.38, 0.12] after fusion feature linear transformation, is solved using the four-section Runge-Kutta method (step size set to 0.01) to obtain the chaotic state C, and then the quantum correlation matrix is ​​constructed. Based on the quantum correlation matrix, the input current is 0.495. The change of the mesenchymal membrane potential is simulated by the Izhikevich neuron model, and the output pulse sequence is used to update the topology matrix. , With both values ​​set to 0.5, the chaotic states of the two nodes are 0.45 and 0.32, respectively, and the quantum correlation matrix is ​​0.99. Therefore, the initial bandwidth allocation result is 0.555. Considering the total bandwidth constraint of 1200Mbps and a single link bandwidth of less than 400, the bandwidth allocation matrix is ​​obtained. ; A multi-objective shortest path algorithm was used to obtain the following candidate paths: Path 1: Development chain → Test chain → Production chain, with a multi-scale topological entropy of 1.538, which is 1.7915 after normalization. The available bandwidth for the Development chain → Test chain is 141.6, the quantum correlation is 0.99, the transmission time is 10ms, and the failure probability is 0.01. The available bandwidth for the Test chain → Production chain is 242.3, the quantum correlation is 0.95, the transmission time is 15ms, and the failure probability is 0.05. The weights for each parameter are 0.2, 0.2, 0.3, and 0.3, respectively. The initial path value is 0.52, and the final path value is 0.93. Path 2: Development chain → Production chain → Test chain. The chain has a multi-scale topological entropy of 1.538. The available bandwidth for the development chain → production chain is 216.8, the quantum correlation is 0.85, the transmission time is 12ms, and the failure probability is 0.15. The available bandwidth for the production chain → test chain is 242.3, the quantum correlation is 0.95, the transmission time is 15ms, and the failure probability is 0.05. The weights of each parameter are 0.2, 0.2, 0.3, and 0.3, respectively. The initial path value is 0.53, and the final path value is 0.95. Based on the final path value, the link bandwidth and path are optimized in real time to address anomalies in the 5G equipment supply chain (such as test chain latency and development chain defects), ensuring the high availability of the production chain. Through testing, the method of this invention reduces test chain latency by 52% in the 5G equipment supply chain, ensuring the delivery of mass-produced 5G equipment.

[0055] It should be noted that the dynamic bandwidth allocation and secure transmission system for multi-chain networks provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be allocated to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be further decomposed or combined. For example, the modules in the above embodiments can be merged into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the various modules or steps and are not considered as an improper limitation of the present invention.

[0056] A dynamic bandwidth allocation and secure transmission system for multi-chain networks according to a second embodiment of the present invention includes the following steps: The data acquisition module is configured to collect data from each chain and inter-chain channels in the multi-chain network as multi-dimensional heterogeneous data; the multi-dimensional heterogeneous data includes topology data, traffic data, and event data. The feature extraction module is configured to extract features from each dimension of the multidimensional heterogeneous data, and then entangle and fuse the extracted features to obtain fused features. Based on the topology data, topology features are extracted using graph isomorphic network (GIN); based on the flow data, a singularity spectrum is obtained using multifractal analysis (MFA); based on the singularity spectrum and the flow data, self-similar features are obtained using MFA; based on the event data, the probability distribution of each type of event is calculated; based on the probability distribution, quantum states are constructed and spliced ​​together to form quantum event states. Entangling and fusing the topological features, the self-similar features, and the quantum event states through quantum gate operations yields the fused features: The quantum event state is converted into a real number vector and then tensor-producted with the topological feature to obtain the first feature; the sum of the first feature and the self-similar feature is calculated and compressed using the ReLU activation function, and the compressed feature is used as the fusion feature; The bandwidth allocation and secure transmission module is configured to obtain a bandwidth allocation matrix and a path planning tensor based on the fusion features and through a trained dynamic bandwidth allocation and secure transmission network; the dynamic bandwidth allocation and secure transmission network is constructed based on a deep neural network.

[0057] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related explanations of the method described above can be found in the corresponding process in the foregoing system embodiments, and will not be repeated here.

[0058] A third embodiment of the present invention provides a dynamic bandwidth allocation and secure transmission device for multi-chain networks, comprising at least one processor and a memory communicatively connected to at least one processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to implement the aforementioned dynamic bandwidth allocation and secure transmission method for multi-chain networks.

[0059] A fourth embodiment of the present invention provides a computer-readable storage medium storing computer instructions, which are executed by the computer to implement the above-described method for dynamic bandwidth allocation and secure transmission in a multi-chain network.

[0060] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the above-described dynamic bandwidth allocation and secure transmission device and computer-readable storage medium for multi-chain networks can be found in the corresponding process in the aforementioned system example, and will not be repeated here.

[0061] Those skilled in the art will recognize that the modules and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. The programs corresponding to the software modules and method steps can be placed in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art. To clearly illustrate the interchangeability of electronic hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the invention.

[0062] The terms “first,” “second,” “third,” etc., are used to distinguish similar objects, not to describe or indicate a specific order or sequence.

[0063] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for dynamic bandwidth allocation and secure transmission in multi-chain networks, characterized in that, The method includes the following steps: S10, Collect data from each chain and inter-chain channels in the multi-chain network as multi-dimensional heterogeneous data; the multi-dimensional heterogeneous data includes topology data, traffic data, and event data; S20, feature extraction is performed on each dimension of the heterogeneous data in the multidimensional heterogeneous data, and the extracted features are entangled and fused to obtain fused features: Based on the topology data, topology features are extracted using graph isomorphic network (GIN); based on the flow data, a singularity spectrum is obtained using multifractal analysis (MFA); based on the singularity spectrum and the flow data, self-similar features are obtained using MFA; based on the event data, the probability distribution of each type of event is calculated; based on the probability distribution, quantum states are constructed and spliced ​​together to form quantum event states. Entangling and fusing the topological features, the self-similar features, and the quantum event states through quantum gate operations yields the fused features: The quantum event state is converted into a real number vector and then tensor-producted with the topological feature to obtain the first feature; the sum of the first feature and the self-similar feature is calculated and compressed using the ReLU activation function, and the compressed feature is used as the fusion feature; S30, based on the fusion features, obtain the bandwidth allocation matrix and path planning tensor through the trained dynamic bandwidth allocation and secure transmission network; the dynamic bandwidth allocation and secure transmission network is constructed based on a deep neural network.

2. The dynamic bandwidth allocation and secure transmission method for multi-chain networks according to claim 1, characterized in that, Based on the aforementioned topological data, topological features are extracted using Graph Isomorphic Network (GIN). The method is as follows: Based on the topology data, the degree of each node in the topology structure is obtained; the degree of two nodes and the communication time interval between the two nodes are concatenated into vectors, and then processed by a multilayer perceptron and an activation function in sequence to obtain the link survival probability between the two nodes. A mask matrix is ​​constructed based on the link survival probability between the two nodes; the Hadamard product between the static connection matrix between the two nodes and the mask matrix is ​​calculated, and then added to the event-driven adjacency matrix between the two nodes to obtain the dynamic adjacency matrix; The degree of each node is concatenated with the encoded vector of the node type information. The concatenated vector is then multiplied with the set parameter matrix and processed by the activation function to obtain the initial feature vector of each node. The initial feature vectors of each node and the dynamic adjacency matrix are input into the GNN network, and the outputs of different layers of the GNN network are fused to obtain fused features. Based on the fusion features, calculate the average feature value of all nodes in each chain of the multi-chain network after MLP processing; The average feature value of other chains is weighted according to the attention weight between the chain to which the current node belongs and other chains, and then added to the fused feature of the current node as the enhanced feature; Historical features are obtained by performing dilated causal convolution on the historical topological data of the current node. The historical features and the enhanced features are then concatenated and combined with a learnable parameter matrix and an activation function to obtain the sensitivity weights of the current node to events. The historical features and the enhanced features are then weighted based on the sensitivity weights to obtain temporal enhanced features, which are the final extracted topological features.

3. The dynamic bandwidth allocation and secure transmission method for multi-chain networks according to claim 1, characterized in that, The bandwidth allocation matrix and path planning tensor are obtained through a trained dynamic bandwidth allocation and secure transmission network. The method is as follows: The fused features are mapped to the chaotic space using a linear transformation as the initial values ​​for the Lorenz differential equation; the Lorenz differential equation is iterated and solved using the four-section Runge-Kutta method to obtain the trajectory point sequence; the state of each node at each time point in the chaotic space is sampled from the trajectory point sequence as the chaotic state C; The chaotic state C is compressed and an entangled state is constructed; based on the entangled state, the quantum mutual information between node pairs is calculated, and then the quantum correlation matrix Q is constructed; Combining the quantum correlation matrix Q, a time-varying input current is constructed through chaotic gating modulation; based on the time-varying input current, the change of the mesenchymal membrane potential is simulated through the Izhikevich neuron model, and a pulse sequence S is output. Based on the pulse sequence S, the pulse energy integral E is calculated; if the pulse energy integral exceeds the set energy threshold, a phase transition is performed on the pulse energy integral, and the phase transition result is elastically embedded and deformed to update the original topology matrix of each node, thereby obtaining the updated topology matrix. Combining the updated topology matrix, the chaotic state, the quantum correlation matrix, and the pulse sequence, the bandwidth allocation matrix and path planning tensor are output.

4. The dynamic bandwidth allocation and secure transmission method for multi-chain networks according to claim 1, characterized in that, The bandwidth allocation matrix is ​​obtained as follows: Based on the updated topology matrix, the chaotic state, and the quantum correlation matrix, the initial bandwidth allocation result is determined; Based on the established physical constraints and the initial bandwidth allocation results, the optimal bandwidth allocation results are calculated using the projection gradient descent method. Then, the bandwidth allocation matrix is ​​obtained by dividing the bandwidth allocation into different levels.

5. The dynamic bandwidth allocation and secure transmission method for multi-chain networks according to claim 1, characterized in that, The path planning tensor is obtained as follows: Based on the bandwidth allocation matrix, the risk warning vector, and the quantum correlation matrix, a multi-objective shortest path algorithm is used to obtain the path planning tensor, i.e., the optimal path or candidate path.

6. The dynamic bandwidth allocation and secure transmission method for multi-chain networks according to claim 5, characterized in that, The path planning tensor is obtained using a multi-objective shortest path algorithm, and the method is as follows: Based on the bandwidth allocation matrix, the risk warning vector, and the quantum correlation matrix, obtain the path available bandwidth, path failure probability, path transmission time, and the minimum value in the quantum correlation matrix corresponding to all edges of each path in the current network; perform a weighted summation of the path available bandwidth, path failure probability, path transmission time, and the minimum value in the quantum correlation matrix corresponding to all edges of each path to obtain the initial path value. Based on the degree of each node in each path, the sum of local degrees, and the sum of the degrees of all nodes in the network, the multi-scale topological entropy corresponding to each path is calculated; the contribution of the topological entropy is determined according to the ratio of each multi-scale topological entropy to the maximum value of the multi-scale topological entropy of all paths in the current network; the sum of local degrees is the sum of the degrees of all nodes in each path. Calculate the similarity between each path and the paths in the selected path set, multiply each similarity by the corresponding time decay factor and sum them up, and take the difference between 1 and the sum as the contribution of the path interaction entropy. The initial path value, the contribution of the topological entropy, and the contribution of the path interaction entropy are multiplied together to obtain the final path value for each path. The path corresponding to the maximum value of the final path value is selected as the optimal path, or the path corresponding to the maximum value of a predetermined number of paths is selected as the candidate path.

7. The method for dynamic bandwidth allocation and secure transmission in multi-chain networks according to claim 1, characterized in that, The entangled fusion of the topological features, the self-similar features, and the quantum event states is achieved through quantum gate operations to obtain fused features. The method is as follows: ； in, Indicates fusion characteristics, Representing topological features, Indicates self-similarity features, Representing quantum event states, This represents the weight vector set during feature fusion.

8. A dynamic bandwidth allocation and secure transmission system for multi-chain networks, employing the dynamic bandwidth allocation and secure transmission method for multi-chain networks as described in any one of claims 1-7, characterized in that, The system includes: The data acquisition module is configured to collect data from each chain and inter-chain channels in the multi-chain network as multi-dimensional heterogeneous data; the multi-dimensional heterogeneous data includes topology data, traffic data, and event data. The feature extraction module is configured to extract features from each dimension of the multidimensional heterogeneous data, and then entangle and fuse the extracted features to obtain fused features. Based on the topology data, topology features are extracted using graph isomorphic network (GIN); based on the flow data, a singularity spectrum is obtained using multifractal analysis (MFA); based on the singularity spectrum and the flow data, self-similar features are obtained using MFA; based on the event data, the probability distribution of each type of event is calculated; based on the probability distribution, quantum states are constructed and spliced ​​together to form quantum event states. Entangling and fusing the topological features, the self-similar features, and the quantum event states through quantum gate operations yields the fused features: The quantum event state is converted into a real number vector and then tensor-producted with the topological feature to obtain the first feature; the sum of the first feature and the self-similar feature is calculated and compressed using the ReLU activation function, and the compressed feature is used as the fusion feature; The bandwidth allocation and secure transmission module is configured to obtain a bandwidth allocation matrix and a path planning tensor based on the fusion features and through a trained dynamic bandwidth allocation and secure transmission network; the dynamic bandwidth allocation and secure transmission network is constructed based on a deep neural network.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement the dynamic bandwidth allocation and secure transmission method for multi-chain networks as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are executed by the computer to implement the dynamic bandwidth allocation and secure transmission method for multi-chain networks as described in any one of claims 1-7.

Images