An enhanced practical byzantine fault tolerance method for service function chaining deployment

Abstract

The application discloses an enhanced practical Byzantine fault tolerance method for service function chain deployment. Firstly, a three-layer trusted network system model integrating blockchain and deep reinforcement learning is constructed, and network parameters and SFC deployment constraints are defined. Then, a VRPBFT enhanced consensus mechanism integrating a verifiable random function (VRF) and a node reputation grading model is designed to quantify node credibility and realize dynamic hierarchical division. Next, a master node fair selection method based on VRF is proposed to reduce consensus delay and improve Byzantine node detection capability. Finally, an SDRL deep reinforcement learning deployment algorithm is designed to dynamically adjust VNF and link deployment strategies in combination with node credibility, thereby optimizing resource utilization and service quality. Compared with the traditional PBFT, the consensus delay is reduced by about 30%, and the proportion of Byzantine nodes is reduced by 40% after 100 rounds of consensus. Compared with existing algorithms, the long-term average income of the SDRL algorithm is increased by 17%, the SFC request acceptance rate is increased by 14.49%, the income-cost ratio is increased by 20.35%, the CPU resource utilization rate is 42% and is increased by 27.96%, the safety of SFC deployment in a heterogeneous network is ensured, and the collaborative optimization of resource utilization and service quality is realized, so that the application is suitable for SFC trusted deployment in a heterogeneous network environment such as the Internet of Things and 5G.

Description

Technical Field

[0001] This invention belongs to the intersection of heterogeneous networks and blockchain technology, specifically involving a trusted deployment method for Service Function Chains (SFC), which is particularly suitable for the secure deployment and resource optimization of SFC in heterogeneous network environments such as the Internet of Things and 5G. Background Technology

[0002] With the rapid development of IoT technology, the interconnectivity of devices in heterogeneous network environments has been greatly improved, giving rise to diverse application scenarios such as smart homes and autonomous driving. Service Function Chain (SFC), as a technology that combines multiple Virtual Network Functions (VNFs) in a specific order, has become a core means to meet users' personalized network service needs and is widely used in scenarios such as traffic monitoring, firewall protection, and intrusion detection.

[0003] However, the high interconnectivity of heterogeneous networks brings numerous security challenges, including data privacy breaches, delegation trust issues, and security risks during SFC deployment. Dynamic issues such as node failures, link congestion, and network attacks (e.g., DoS attacks, node forgery, request tampering) seriously threaten the reliability of SFCs, leading to compromised network security, communication quality, and user experience. Furthermore, selecting the optimal deployment path and nodes for SFCs in complex network topologies to meet Quality of Service (QoS) requirements such as low latency, high bandwidth, and high reliability becomes a complex optimization challenge.

[0004] Traditional Practical Byzantine Fault Tolerance (PBFT) consensus mechanisms suffer from high consensus latency and low Byzantine node detection efficiency, making them unsuitable for the dynamic deployment requirements of SFC (System-Driven Fault). Existing SFC deployment algorithms often ignore dynamic changes in node trustworthiness, resulting in limited resource utilization and service quality optimization, failing to achieve efficient SFC deployment while ensuring security. Therefore, there is an urgent need to design an enhanced PBFT mechanism that combines blockchain and deep reinforcement learning technologies to achieve reliable and efficient SFC deployment. Summary of the Invention

[0005] The purpose of this invention is to solve the security and resource efficiency problems of SFC deployment in heterogeneous networks. It provides an enhanced practical Byzantine fault-tolerant method for service function chain deployment, which integrates verifiable random functions (VRF), node reputation rating models, blockchain and deep reinforcement learning to reduce consensus latency, improve Byzantine node detection capabilities, and optimize resource utilization, service quality and security of SFC deployment.

[0006] The technical solution to achieve the purpose of this invention is: an enhanced practical Byzantine fault-tolerant method for service function chain deployment, comprising the following steps:

[0007] Step 1: Construct a trusted network system model that integrates blockchain and deep reinforcement learning, including a three-layer architecture of heterogeneous network physical layer, virtual network layer and blockchain layer, and define network model parameters and deployment constraints.

[0008] Step 2: Design an enhanced PBFT consensus mechanism (VRPBFT) that integrates a verifiable random function (VRF) and a node reputation rating model to quantify node credibility and realize node hierarchy partitioning and dynamic transfer;

[0009] Step 3: Propose a master node selection method based on VRPBFT, which completes the fair selection of master nodes through VRF key pair generation, random number verification and sorting;

[0010] Step 4: Design a deep reinforcement learning-based SFC deployment algorithm (SDRL), dynamically adjust the deployment strategy based on node credibility, and implement virtual link deployment through the shortest path algorithm to optimize resource utilization and service quality.

[0011] Furthermore, the construction of the trusted network system model described in step 1 is as follows:

[0012] Model the physical layer of heterogeneous networks as a weighted undirected graph. ,in For the CPU resources of physical nodes, For node energy capacity, For node security level, Physical link bandwidth resources; the operating probability of physical nodes and links follows an interval. Uniform distribution within.

[0013] Set the security level constraints for SFC deployment as follows: ,in For the security level requirements of virtual nodes, To ensure the trustworthiness of physical nodes, the security level of physical nodes must meet the deployment requirements of virtual nodes.

[0014] The algorithm evaluation metrics include consensus latency, long-term average revenue, revenue-cost ratio, SFC request acceptance rate, and CPU resource utilization.

[0015] 1. Consensus delay:

[0016] ,

[0017] in For the request generation time, The time for consensus confirmation;

[0018] 2. Long-term average return:

[0019]

[0020] 3. Deployment costs:

[0021] ;

[0022] 4. Long-run average benefit-cost ratio:

[0023] .

[0024] The blockchain layer is responsible for the registration and updating of resources and transactions during the SFC deployment process. All physical node information is registered to the blockchain before the service is provided to ensure the data security of VNF node deployment. The deep reinforcement learning module determines the optimal SFC deployment strategy based on the node resources, location and trust data provided by the blockchain.

[0025] Furthermore, the design of the VRPBFT consensus mechanism described in step 2 is as follows:

[0026] Node credibility It is a comprehensive evaluation index for the interval [0,1], used to measure the performance of a node in the th interval. The credibility level in the consensus round is calculated as follows:

[0027]

[0028] in:

[0029] 1. Cost ratio for nodes For nodes The deposit investment, Total deposit;

[0030] 2. For node communication success rate, For nodes The number of times communication was successfully completed. For nodes Total number of communication requests;

[0031] 3. For node honesty rate, For nodes The number of acts of honesty This represents the total number of times a node participates in the consensus process.

[0032] 4. To assess the historical reliability of nodes;

[0033] 5. Each indicator is assigned a weight and summed to 1. This can be flexibly adjusted according to different application scenarios to balance the contribution of each indicator in the node credibility assessment.

[0034] The node credibility is calculated and updated through a node credibility evaluation algorithm. The algorithm takes the cardinality of the node set and the historical node credibility registry as inputs and outputs a new node credibility value record table. The algorithm calculates the input cost rate, communication success rate and honesty rate of each node in turn, and combines them with the historical credibility to obtain the node credibility, and updates the credibility registry in real time.

[0035] Nodes are categorized into four levels of credibility: A, B, C, and D. The credibility range for category A nodes is... Class B nodes are high-reputation nodes and can serve as priority master nodes, participate in consensus, and verify random numbers; the credibility range of Class B nodes is... Class C nodes are considered high-reputation nodes and can act as master nodes and participate in consensus, but they lack full verification permissions; the credibility range of Class C nodes is... Class D nodes are medium-reputation nodes, participating only in consensus with limited permissions; the credibility range of Class D nodes is... Nodes with low credibility are excluded from the consensus process and require strict verification. Node hierarchy is dynamically adjusted, with credibility updated in real-time based on node behavior, ensuring network security and efficiency.

[0036] Furthermore, the master node selection method described in step 3 specifically includes the following steps:

[0037] 1. Key pair generation: All honest nodes of type A and B generate their own VRF key pairs. The private key is kept by the node and used to generate random numbers, while the public key is distributed to other nodes for verification of the random numbers.

[0038] 2. Public Key Distribution: Nodes broadcast their VRF public keys to the entire network, ensuring that all consensus nodes can obtain them for subsequent random number verification;

[0039] 3. Random number generation: The node combines its own VRF private key with the current timestamp to generate a unique random number. The timestamp ensures the timeliness and uniqueness of the random number.

[0040] 4. Random number broadcasting: The node broadcasts the generated random number, its own VRF public key, and timestamp to the network for other nodes to verify;

[0041] 5. Random number verification: Consensus nodes in the network verify the validity of random numbers by using the broadcast VRF public key and timestamp, and discard invalid random numbers;

[0042] 6. Random Number Sorting: Collect all valid random numbers and sort them in ascending order of their values;

[0043] 7. Master Node Selection: Select the node with the smallest random number as the master node; if multiple nodes have the same random number, select the node with higher credibility as the master node to ensure the fairness and randomness of the master node selection.

[0044] Furthermore, the design of the SDRL deployment algorithm described in step 4 is as follows:

[0045] The SDRL algorithm uses physical networks Virtual networks Set the iteration rounds as input. Learning rate Batch size It outputs the node embedding probability and dynamically optimizes the SFC deployment strategy based on node trustworthiness.

[0046] Link deployment is achieved through the shortest path algorithm. The specific steps are as follows: for each virtual link, the top 5 shortest paths in the physical network are extracted, and paths that do not meet the resource and security level constraints are eliminated. If a valid path exists, one of them is selected to complete the link deployment, and the available CPU, bandwidth and other resources of the physical network are updated in real time. If there is no valid path, the link deployment fails.

[0047] Compared with the prior art, the significant advantages of this invention are:

[0048] (1) A VRPBFT enhanced consensus mechanism was designed, which integrates VRF and node reputation rating model. Compared with traditional PBFT, the consensus latency is reduced by about 30%, and the proportion of Byzantine nodes is reduced by 40% after 100 rounds of consensus, which greatly improves the consensus efficiency and Byzantine node detection capability.

[0049] (2) A dynamic hierarchical and transfer mechanism based on node credibility is proposed. The reputation is evaluated in real time according to the node behavior, and the node permissions are dynamically adjusted to ensure the security and reliability of the network.

[0050] (3) An SDRL deep reinforcement learning deployment algorithm was designed, and the deployment strategy was dynamically adjusted in combination with node credibility. Compared with the existing algorithm, the long-term average benefit was increased by 17%, the SFC request acceptance rate was increased by 14.49%, the benefit-cost ratio was increased by 20.35%, and the CPU resource utilization rate reached 42%, an improvement of 27.96%, realizing the synergistic optimization of resource utilization, service quality and security.

[0051] (4) A trusted network system integrating blockchain and deep reinforcement learning was constructed. Blockchain enables secure storage of node information and transaction traceability, while deep reinforcement learning enables intelligent decision-making for SFC deployment, providing an efficient solution for trusted deployment of SFC in heterogeneous networks. Attached Figure Description

[0052] Figure 1 is a diagram of the trusted network system architecture in this invention;

[0053] Figure 2 is a schematic diagram of the working principle of the trusted network layer in this invention;

[0054] Figure 3 is a schematic diagram of the node hierarchy in this invention. Detailed Implementation

[0055] The present invention will now be described in detail with reference to the accompanying drawings. The present invention discloses an enhanced practical Byzantine fault-tolerant method for service function chain deployment, comprising the following steps:

[0056] Step 1: Construct a trusted network system model that integrates blockchain and deep reinforcement learning, as follows:

[0057] A three-layer trusted network system architecture is constructed, consisting of a heterogeneous network physical layer, a virtual network layer, and a blockchain layer. The physical layer provides physical nodes and link resources for SFC deployment, the virtual layer provides users with personalized VNF combination services, and the blockchain layer is responsible for resource registration, transaction updates, and security authentication.

[0058] Model the physical layer of heterogeneous networks as a weighted undirected graph. Set the number of physical nodes to 100 and the number of physical links to 530. Divide the physical nodes into three categories: X, Y, and Z. There are 10 nodes in category X. CPU resources are allocated according to... Uniformly distributed; 25 Y-type nodes, CPU resources follow... Uniformly distributed; 65 Z-type nodes, CPU resources follow... Uniformly distributed; physical links are divided into X, Y, Z classes and inter-domain links, with X link bandwidth... Mbps, Y link Mbps, Z-link Mbps, inter-domain link Mbps; the operating probability of physical nodes and links follows an interval Uniform distribution within.

[0059] Set the security level constraints for SFC deployment as follows: The security requirements for virtual nodes are divided into four levels: A, B, C, and D. The trustworthiness of physical nodes must be higher than that of virtual nodes.

[0060] The algorithm evaluation metrics include consensus latency, long-term average revenue, revenue-cost ratio, SFC request acceptance rate, and CPU resource utilization, among which consensus latency... Long-run average benefit-cost ratio .

[0061] Step 2: Design an enhanced PBFT consensus mechanism (VRPBFT) that integrates a verifiable random function (VRF) and a node reputation rating model, as detailed below:

[0062] Define node credibility Set the weights for each indicator The summation is 1; calculate the input cost rate of each node sequentially. Honesty rate Combining historical credibility The node credibility is obtained, and the node credibility range is [0,1].

[0063] The node credibility is updated through a node credibility evaluation algorithm. The input node set cardinality is 100 and the historical node credibility registry. The policy network parameters and node credibility list are initialized. All nodes are traversed, the historical data of nodes are read, the values ​​of various indicators and node credibility are calculated, the historical node credibility registry is updated, and a new node credibility value record table is output.

[0064] Step 3: Propose a master node selection method based on VRPBFT. The specific steps are as follows:

[0065] 1. Key Pair Generation: All honest nodes of type A and B generate VRF key pairs, with the private key stored locally and the public key distributed externally;

[0066] 2. Public Key Distribution: Nodes broadcast their VRF public keys to the network, ensuring that all A, B, and C class consensus nodes can obtain them;

[0067] 3. Random number generation: The node combines the VRF private key with the current timestamp to generate a unique random number;

[0068] 4. Random number broadcasting: Nodes broadcast random numbers, VRF public keys, and timestamps;

[0069] 5. Random number verification: Consensus nodes verify the validity of random numbers using public keys and timestamps, and discard invalid values;

[0070] 6. Random Number Sorting: Sort valid random numbers in ascending order;

[0071] 7. Master Node Selection: Select the node with the smallest random number as the master node. If the random numbers are the same, select the node with higher reliability.

[0072] Step 4: Design the SFC deployment algorithm (SDRL) based on deep reinforcement learning, as follows:

[0073] 2000 SFC requests are generated, of which 1000 are used for training and 1000 for testing; the number of VNFs in each SFC request follows a uniform distribution of U[2,10], the CPU resource requirement of VNFs is U[3,50] cores, the bandwidth resource requirement is U[3,50] Mbps, and the security level requirement is A, B, C, or D; the arrival time of SFC requests follows a Poisson distribution, with an average of 4 requests arriving every 100 time units.

[0074] The algorithm execution process is as follows:

[0075] 1. Initialization: Randomly initialize the policy network parameters, initialize the node reputation list, and record the initial credibility of 100 physical nodes;

[0076] 2. Iterative training: Loop through 200 rounds, iterating through 1000 training set SFC requests;

[0077] 3. Feature Extraction: Extract features such as CPU, bandwidth, security level, and trustworthiness of physical nodes to generate a feature matrix;

[0078] 4. Probability Generation: Generate the VNF embedding probability of each node through a policy network;

[0079] 5. Trusted Node Filtering: Pseudo-random values ​​are generated using VRF, and a reputation threshold of 0.3 is set to filter out A, B, and C class nodes with a trustworthiness ≥ 0.3;

[0080] 6. VNF Embedding: Based on the embedding probability, select nodes from trusted nodes to complete VNF deployment;

[0081] 7. Link Embedding: If all VNFs are mapped, the shortest path algorithm is used to extract the top 5 shortest physical paths, paths that do not meet the constraints are eliminated, valid paths are selected to complete the virtual link deployment, and the available physical network resources are updated;

[0082] 8. Parameter Update: Calculate the loss value, reward value, and gradient, update the agent parameters through backpropagation, and adjust the node credibility based on the reward value of the deployment effect; if the VNF mapping is not completed, clean up the gradient and continue iterating.

[0083] The experiment uses NetworkX tools to build a heterogeneous network topology and Hyperledger Fabric to implement a blockchain network. It compares the SDRL algorithm with traditional RL algorithms and SA-RL algorithms, and the VRPBFT mechanism with the traditional PBFT mechanism.

[0084] Experimental results show that VRPBFT reduces consensus latency by about 30% compared to traditional PBFT, and reduces the proportion of Byzantine nodes by 40% after 100 rounds of consensus. Compared with existing algorithms, the SDRL algorithm improves long-term average returns by 17%, SFC request acceptance rate by 14.49%, benefit-cost ratio by 20.35%, and CPU resource utilization by 42%, an improvement of 27.96%, achieving coordinated optimization of security and efficiency in SFC deployment.

Claims

1. An enhanced practical Byzantine fault-tolerant method for service function chain deployment, characterized in that, Includes the following steps: Step 1: Construct a trusted network system model that integrates blockchain and deep reinforcement learning, including a three-layer architecture of heterogeneous network physical layer, virtual network layer and blockchain layer, and define network model parameters and deployment constraints. Step 2: Design an enhanced PBFT consensus mechanism (VRPBFT) that integrates a verifiable random function (VRF) and a node reputation rating model to quantify node credibility and realize node hierarchy partitioning and dynamic transfer; Step 3: Propose a master node selection method based on VRPBFT, which completes the fair selection of master nodes through VRF key pair generation, random number verification and sorting; Step 4: Design a deep reinforcement learning-based SFC deployment algorithm (SDRL), dynamically adjust the deployment strategy based on node credibility, and implement virtual link deployment through the shortest path algorithm to optimize resource utilization and service quality.

2. The enhanced practical Byzantine fault-tolerant method for service-oriented function chain deployment according to claim 1, characterized in that, Step 1, which involves constructing a trusted network system model, is as follows: Model the physical layer of heterogeneous networks as a weighted undirected graph. ,in For the CPU resources of physical nodes, For node energy capacity, For node security level, Physical link bandwidth resources; the operating probability of physical nodes and links follows an interval. Uniform distribution within; Set the security level constraints for SFC deployment as follows: ,in For the security level requirements of virtual nodes, The trustworthiness of physical nodes; Define algorithm evaluation metrics including consensus latency Long-term average revenue, revenue-cost ratio, SFC request acceptance rate, and CPU resource utilization. Among the profits , cost , Long-run average benefit-cost ratio .

3. The enhanced practical Byzantine fault-tolerant method for service-oriented function chain deployment according to claim 1, characterized in that, Step 2 describes the design of the VRPBFT consensus mechanism as follows: Node credibility The comprehensive evaluation index for the interval [0,1] is calculated as follows: in Cost ratio for nodes For node communication success rate, For node honesty rate, To ensure the historical reliability of nodes, Each indicator has a weight, and the sum of these weights is 1. Nodes are categorized into four levels of credibility: A, B, C, and D. The credibility range for category A nodes is... Category B is Class C is Class D is ; The node hierarchy is dynamically transferred. Class A and B can serve as master nodes, Class A, B, and C are consensus nodes, and Class D is excluded from consensus.

4. The enhanced practical Byzantine fault-tolerant method for service-oriented function chain deployment according to claim 1, characterized in that, The master node selection method described in step 3 consists of the following steps: (1) All honest nodes generate their own VRF key pairs, with the private key used to generate random numbers and the public key distributed to the outside for verification; (2) The node generates a random number by combining the VRF private key and the current timestamp, and broadcasts the random number, VRF public key and timestamp to the network; (3) Other nodes in the network verify the validity of the random number by using the broadcast public key and timestamp; (4) Collect all valid random numbers and sort them by size, and select the node with the smallest random number as the master node; If the random numbers are the same, the node with higher credibility is selected as the master node.

5. The enhanced practical Byzantine fault-tolerant method for service-oriented function chain deployment according to claim 1, characterized in that, Step 4 describes the design of the SDRL deployment algorithm, as follows: The SDRL algorithm uses physical networks Virtual networks For input, iteration rounds Learning rate Batch size Output the node embedding probability; The algorithm execution steps are as follows: randomly initialize policy network parameters and node reputation list, traverse SFC requests in the training set, extract node feature matrix and generate deployment probability, generate pseudo-random values ​​through VRF and filter trusted nodes in combination with reputation list, select nodes based on probability to complete virtual network function (VNF) embedding, if all VNFs are mapped, perform link embedding, calculate reward and gradient and backpropagate to update agent parameters, and dynamically adjust node reputation according to reward value; Link deployment is achieved through the shortest path algorithm, which extracts the top k shortest paths and removes paths that do not meet the constraints, selects the valid paths to complete the deployment, and updates the available physical network resources. The value of k is 5.

Images