Industrial internet of things slice security service configuration method based on deep reinforcement learning

Abstract

The application provides an industrial Internet of Things slice security service configuration method based on deep reinforcement learning. The industrial Internet of Things slice security service configuration method based on deep reinforcement learning comprises the following steps: building an industrial Internet of Things architecture, wherein the industrial Internet of Things architecture comprises a terminal device, a network slice module, a virtual security network function module, an arrangement module and a physical network; in the virtual security network function module, based on software-defined network and network function virtualization technology, a security service request model is obtained by modeling a security service request of a user; a security service function chain deployment scheme is obtained by model training based on a security service network function instance and a security service function chain; and the security service function chain is embedded into the industrial Internet of Things architecture based on slice characteristics of the network slice, request distribution of the network slice and a simplified action space. The method can formulate an effective security policy and a resource optimization scheme when facing a large number of network slices and service requests.

Description

Technical Field

[0001] This invention relates to the field of communication network technology, and in particular to a method for configuring secure services for industrial IoT slices based on deep reinforcement learning. Background Technology

[0002] The Industrial Internet of Things (IIoT) integrates various sensors and controllers with sensing and monitoring capabilities, along with technologies such as mobile communication and intelligent analytics, into all aspects of industrial production processes. This significantly improves manufacturing efficiency, enhances product quality, reduces product costs and resource consumption, ultimately elevating traditional industries to a new stage of intelligent manufacturing. In terms of application, IIoT is characterized by real-time performance, automation, embedded (software) integration, security, and interconnectivity. IIoT has wide applications in smart manufacturing, healthcare, agriculture, and transportation.

[0003] Network slicing technology provides diverse functional services to various industrial entities to meet the diverse and dynamic application needs of latency-intensive Industrial Internet of Things (IIoT). As a novel network architecture technology, network slicing offers a high degree of control and flexibility over network resources, enabling customized network services to be provided based on different business requirements.

[0004] In existing technologies, due to the complexity and dynamism of network slicing environments, traditional security strategies and resource management methods often fall short of meeting the demands. In particular, when faced with a large number of network slices and diverse business requests, developing effective security strategies and resource optimization solutions has become a significant challenge. Summary of the Invention

[0005] In view of this, the present invention provides a method for configuring security services for industrial IoT slicing based on deep reinforcement learning to solve the above problems.

[0006] According to a first aspect of the present invention, a method for configuring secure services for industrial IoT slices based on deep reinforcement learning is provided, comprising: constructing an industrial IoT architecture, the industrial IoT architecture including terminal devices, network slicing modules, virtual secure network function modules, orchestration modules, and physical networks; in the virtual secure network function modules, modeling user security service requests based on software-defined networking and network function virtualization technologies to obtain a security service request model; training the model based on security service network function instances and security service function chains to obtain a security service function chain deployment scheme; and embedding the security service function chain into the industrial IoT architecture based on the slice characteristics of network slices, the request distribution of network slices, and a simplified action space.

[0007] In another implementation of the present invention, in the industrial Internet of Things architecture, different network slices have different distributions; the network slice module adjusts the distribution of network slices based on slice characteristics.

[0008] In another implementation of the present invention, the security service request model includes a virtualized security service function chain, multi-dimensional security levels, and resource pricing factors.

[0009] In another implementation of the present invention, the method further includes: the terminal device sending a security service request to the network operator; the network operator determining a security service network function instance mapped to the corresponding virtual security network service function based on the security service function chain deployment scheme; and satisfying the security service function chain deployment scheme through a service satisfaction index, wherein the service satisfaction index includes security level and service resources.

[0010] In another implementation of the present invention, the security level is calculated as follows: based on the security service network function instance and its corresponding virtual security network function, the relationship between its expected security level and the actual security level is compared; an indicator function is used to indicate whether each correspondence meets the security level, and if the expected security level value is greater than or equal to the actual security level value, it indicates that the correspondence meets the security level.

[0011] In another implementation of the present invention, resource pricing factors include computing resources and bandwidth resources; service resources are calculated by using a utility function to represent satisfaction with computing resources and bandwidth resources.

[0012] In another implementation of the present invention, a security service function chain is embedded into an industrial Internet of Things (IoT) architecture based on the slice features of network slices, the request distribution of network slices, and a simplified action space. This includes: extracting slice features from network slices using a graph neural network; learning and predicting using a GRU model to obtain the request distribution of network slices; simplifying the action space using a sequence-to-sequence model; and embedding the security service function chain into the industrial IoT architecture based on the slice features, request distribution, and simplified action space.

[0013] In the deep reinforcement learning-based industrial IoT slicing security service configuration method of this invention, the security and resource requirements of different network slices are understood and met through a security service request model. The security service function chain deployment scheme is met through Quality of Service (QoS), and the security level and resource provision are evaluated and met to ensure that the quality of service is met. The GS2L method is used to effectively handle the high-dimensional security service function chain (SSFC) embedding and resource allocation problem, so as to provide efficient and security-compliant services for different types of services. This enables the formulation of effective security strategies and resource optimization schemes when faced with a large number of network slices and service requests. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. By reading the detailed description of the embodiments below, the advantages and benefits of the solutions will become clear to those skilled in the art. The accompanying drawings are only for illustrating preferred embodiments and are not intended to limit the present invention.

[0015] In the picture:

[0016] Figure 1 This is a schematic diagram of the industrial Internet of Things architecture according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the workflow under the industrial Internet of Things architecture of another embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of a GS2L reinforcement learning architecture based on graph neural networks and sequences, according to another embodiment of the present invention.

[0019] Figure 4 This is a schematic diagram of the Fat-tree architecture used in the simulation of another embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram illustrating the rewards of GS2L under different learning rates, representing another embodiment of the present invention.

[0021] Figure 6 This is a schematic diagram of the rewards of GS2L under different network slices, which is another embodiment of the present invention.

[0022] Figure 7 This is a schematic diagram illustrating the rewards of different algorithms under different network slices, representing another embodiment of the present invention.

[0023] Figure 8 This is a schematic diagram illustrating the bandwidth resource consumption of different algorithms under different security service requests in network slice 1, according to another embodiment of the present invention.

[0024] Figure 9 This is a schematic diagram illustrating the computational resource consumption of different algorithms under different security service requests in network slice 1, according to another embodiment of the present invention.

[0025] Figure 10 This is a schematic diagram illustrating the QoS metrics of different algorithms under different security service requests in network slice 1, according to another embodiment of the present invention. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art should fall within the protection scope of the present invention.

[0027] Figure 1 This is a schematic diagram of an industrial Internet of Things (IoT) architecture provided as an embodiment of the present invention. For example... Figure 1 As shown, this embodiment mainly includes:

[0028] The industrial IoT architecture is built, which includes terminal devices, network slicing modules, virtual secure network functional modules, orchestration modules, and physical networks.

[0029] For example, build such Figure 1 The illustrated Industrial Internet of Things (SSIOT) architecture addresses the need to provide diverse security services for end devices in an Industrial Internet of Things (IIoT) environment. The SSIOT architecture includes components such as end devices, network slicing modules, VSNF modules, orchestration modules, and physical networks.

[0030] It should be understood that the SSIOT architecture uses Software-Defined Networking (SDN) and Network Functions Virtualization (NFV) technologies to achieve flexible management and deployment of network slices and security services. SDN technology enables network slicing modules to provide different slice network functions based on the logical function of traffic and the utilization of physical resources, while NFV technology allows security service functions to be virtualized into Virtual Security Network Functions (VSNFs), thereby enabling flexible deployment and resource allocation of security services.

[0031] In the virtual secure network function module, based on software-defined networking and network function virtualization technologies, user security service requests are modeled to obtain a security service request model.

[0032] For example, the VSNF module employs a security service request model based on SDN and NFV to model security service requests submitted by clients. This model includes elements such as the virtualized Security Service Function Chain (VSSC), multi-dimensional security level requirements, computing resources, bandwidth resources, and resource pricing factors, effectively understanding and meeting the security and resource requirements of different network slices.

[0033] The model is trained based on security service network function instances and security service function chains to obtain a security service function chain deployment scheme.

[0034] For example, the Security Service Function Chain deployment scheme includes a Security Service Network Function Instance (SNFI) and a Security Service Function Chain (SSFC). This model ensures service quality is met by mapping virtual security service functions to physical network function instances and by using Quality of Service (QoS) to satisfy the assessment and requirements for security levels and resources in the model.

[0035] Based on the slice characteristics of network slices, the request distribution of network slices, and the simplified action space, the security service function chain is embedded into the industrial IoT architecture.

[0036] For example, an AI-based GS2L scheme includes extracting network features using graph neural networks (GNNs), learning slice-specific request distributions using GRUs, and simplifying the action space using a sequence-to-sequence (Seq2Seq) model.

[0037] It should be understood that graph neural networks are used to extract important features and patterns from the network. These features and patterns, including network topology, traffic distribution, and service demands, are encoded as attributes of network nodes and edges, providing an information basis for subsequent policy decisions. The GRU learning model is primarily used to learn and predict the request distribution of network slices. By learning from historical data, the model can predict the possible distribution of different types of service requests over a future period, thereby enabling proactive resource allocation. The sequence-to-sequence (Seq2Seq) model is mainly used to simplify the action space, that is, to learn a policy from historical state and action sequences, allowing the system to choose the optimal action for decision-making in future states. These three parts together constitute the GS2L scheme, enabling automatic security policy formulation and resource optimization based on the distribution of service requests in a network slicing environment.

[0038] In the deep reinforcement learning-based industrial IoT slicing security service configuration method of this invention, the security and resource requirements of different network slices are understood and met through a security service request model. The security service function chain deployment scheme is met through Quality of Service (QoS), and the security level and resource provision are evaluated and met to ensure that the quality of service is met. The GS2L method is used to effectively handle the high-dimensional security service function chain (SSFC) embedding and resource allocation problem, so as to provide efficient and security-compliant services for different types of services. This enables the formulation of effective security strategies and resource optimization schemes when faced with a large number of network slices and service requests.

[0039] In another implementation of the present invention, in the industrial Internet of Things architecture, different network slices have different distributions; the network slice module adjusts the distribution of network slices based on slice characteristics.

[0040] For example, in the design of the SSIOT architecture, different network slices have different distributions, which allows each network slice to optimize itself according to its specific security requirements and resource distribution, thereby enabling customized services to be provided for different types of businesses.

[0041] It should be understood that within the corresponding network slices, specific distribution patterns are implemented based on different service types and security requirements. These patterns include the network slicing module's ability to distribute slices specifically for specific service requests. For example, for services with high real-time requirements, more densely distributed network nodes are used to meet low-latency needs; for services with high security requirements, distribution occurs on nodes with high-level security protection. During the embedding of the Security Service Function Chain (SSFC), specific distribution patterns are implemented based on different service types and security requirements. For example, for services requiring high-level security protection, Security Service Network Function Instances (SNFI) are distributed on physical nodes with high-level security protection capabilities. Based on changes in service traffic and resource demands, the network slicing module dynamically adjusts the distribution of network slices to optimize resource utilization, ensure Quality of Service (QoS) satisfaction, and meet slice-specific requirements, such as maximum tolerable time and computational energy consumption.

[0042] In another implementation of the present invention, the security service request model includes a virtualized security service function chain, multi-dimensional security levels, and resource pricing factors.

[0043] In another implementation of the present invention, the method further includes: the terminal device sending a security service request to the network operator; the network operator determining a security service network function instance mapped to the corresponding virtual security network service function based on the security service function chain deployment scheme; and satisfying the security service function chain deployment scheme through a service satisfaction index, wherein the service satisfaction index includes security level and service resources.

[0044] For example, such as Figure 2 As shown, terminal devices generate security service requests with resource constraints in different slices and report this information to the SDN controller. The SDN controller (network operator) determines which Security Service Network Function Instances (SNFIs) should be mapped to their respective Virtual Security Network Service Functions (VSNFs) based on the security service request model. Finally, based on the Quality of Service (QoS) satisfaction model, it evaluates and satisfies the security level and resource requirements, and embeds the Security Service Function Chain (SSFC). This will consume node and link resources in the network slice.

[0045] In another implementation of the present invention, the security level is calculated as follows: based on the security service network function instance and its corresponding virtual security network function, the relationship between its expected security level and the actual security level is compared; an indicator function is used to indicate whether each correspondence meets the security level, and if the expected security level value is greater than or equal to the actual security level value, it indicates that the correspondence meets the security level.

[0046] For example, the security level QoS is used to assess and meet the security requirements of different network slices. The security level is based on a comparison between the security level of the Security Service Function Chain (SSFC) and the actual security level provided.

[0047] Security Quality of Service (QoS) can be measured and calculated in the following ways:

[0048] For each Security Service Network Function Instance (SNFI) and Virtual Security Network Function (VSNF) pair, compare the relationship between its expected security level (Ple) and the actual security level provided (Rle).

[0049] Indicator functions are used to indicate whether each correspondence is satisfied. If the Ple value of each factor is greater than or equal to the corresponding Rle value, it means that the QoS security level is satisfied.

[0050] In another implementation of the present invention, resource pricing factors include computing resources and bandwidth resources; service resources are calculated by using a utility function to represent satisfaction with computing resources and bandwidth resources.

[0051] For example, QoS is used to assess and manage the resource consumption and allocation of the security service function chain. Resource provision takes into account the relationship between the demand for computing and bandwidth resources and the actual availability.

[0052] Resource provision QoS can be measured and calculated in the following ways:

[0053] The pricing of resources for different network slice types takes into account factors including computing resources and bandwidth resources.

[0054] A utility function is used to represent satisfaction with computing and bandwidth resources. The specific form of the utility function can be designed according to actual needs, such as a linear function or other functional forms, to indicate whether the required resources are met.

[0055] Resource provision QoS is calculated by weighting the satisfaction levels of resource and bandwidth resources. The weights are designed to take into account resource pricing factors and vary depending on the slice. By comprehensively considering security level QoS and resource provision QoS, an overall QoS metric can be obtained, which is used to evaluate and measure the satisfaction level of the security service function chain.

[0056] In another implementation of the present invention, a security service function chain is embedded into an industrial Internet of Things (IoT) architecture based on the slice features of network slices, the request distribution of network slices, and a simplified action space. This includes: extracting slice features from network slices using a graph neural network; learning and predicting using a GRU model to obtain the request distribution of network slices; simplifying the action space using a sequence-to-sequence model; and embedding the security service function chain into the industrial IoT architecture based on the slice features, request distribution, and simplified action space.

[0057] For example, the entire security policy design and resource allocation problem can be modeled as a mixed-integer nonlinear programming (MINLP) problem, the goal of which is to maximize the long-term QoS satisfaction index, i.e., to select the service function chain for each SSR to achieve this goal. The constraints in the established MINLP problem are as follows:

[0058] 1. Ensure that the total bandwidth used by all service function instances does not exceed the capacity of the physical link resources in network slice S;

[0059] 2. Ensure that the total computing resources used by all service function instances in each small data center (MDC) do not exceed the total resource capacity of that MDC;

[0060] 3. Ensure that the service function instances in each SSFC are embedded in the correct order;

[0061] 4. Ensure that the bandwidth and computing resources provided do not exceed the required resources.

[0062] The modeled MINLP problem is an NP-hard problem that cannot be solved in polynomial time. Considering the complexity of the problem and the dynamic nature of the scenario, a deep reinforcement learning-based approach, namely GS2L (GNN-enhanced sequence-to-sequence DRL approach), is designed to address the Security Service Function Chain (SSFC) embedding problem and the resource allocation problem.

[0063] like Figure 3 As shown, the GS2L method achieves embedding and resource scheduling of SSFC by using graph neural networks (GNN) to extract features of network slices, using a GRU model to learn slice-specific request distributions, and simplifying the action space through a sequence-to-sequence (Seq2Seq) model.

[0064] Specifically, the implementation steps of the GS2L method are as follows:

[0065] Feature extraction: First, graph neural networks (GNNs) are used to extract features from the topology and resource allocation of network slices. GNNs can capture the relationships between nodes and edges in the network and extract useful feature representations.

[0066] Request distribution learning: The GRU model is used to model and learn the request distribution specific to a slice. Through the GRU model, different types of business requests can be modeled and decisions can be made based on the specific needs and environmental conditions of the slice.

[0067] Action space simplification: The action space is simplified through a sequence-to-sequence (Seq2Seq) model, which maps a complex action space to a simpler representation, reducing the complexity of decision-making.

[0068] SSFC Embedding and Resource Scheduling: Based on learned feature representations, request distributions, and a simplified action space, the GS2L method is used to embed and schedule resources for SSFCs. By generating sequences of SSFCs and combining them with deep reinforcement learning algorithms, the GS2L method can quickly and accurately generate slice-specific security policy designs and resource scheduling algorithms that meet security requirements in real-time decision-making.

[0069] Through the above implementation methods, this invention can effectively solve the embedding and resource allocation problems of SSFC. The application of the GS2L method enables network administrators to flexibly design security policies and schedule resources according to different network slicing requirements, thereby providing efficient services that meet security needs.

[0070] In another implementation of the present invention, the resource consumption and performance indicators of different schemes are compared in different network slices. Figure 4 A schematic diagram of the Fat-tree architecture used in the simulation, as shown below. Figure 4 As shown, under this architecture, the resource consumption and performance metrics of different schemes in different network slices are compared. Simulations were performed on four network slices with a fixed total of 100 nodes. The simulation comparison schemes include the following:

[0071] Model-NSX: Model-NSX refers to a model trained on a specific network slice, where X takes the value 1, 2, 3, or 4, representing the identifier of the network slice. Specifically, three types of models are provided: GS2L-NSX, GSAC-NSX, and SAC-NSX.

[0072] GS2L-NSAll: GS2L-NSAll refers to a GS2L model trained on all network slices. During training, the GS2L-NSAll model randomly selects SSR embeddings from four network slices and uses them as experience to learn the policy.

[0073] GSAC-NSAll: GSAC-NSAll is a simplified version of the GS2L model, with the GRU network removed. Testing the GSAC model can be seen as an experiment with the GRU network removed to demonstrate the performance of the GRU network within the GS2L model. Similarly, GSAC-NSAll refers to the GSAC model trained on all four network slices.

[0074] SAC-NSAll: SAC-NSAll is also a simplified version of GS2L, which removes the GCN and GRU networks.

[0075] PSO: PSO is a heuristic algorithm that requires no training. It is a commonly used algorithm that has been applied to many service chain embedding problems to solve optimization problems.

[0076] like Figure 5 As shown, the learning rate combination (0.005, 0.005) performed best during training, therefore this parameter setting was chosen as the standard for subsequent training. By adjusting the learning rate, the convergence speed and stability of the model during training can be controlled. A lower learning rate may lead to slower convergence, while a higher learning rate may lead to unstable training. Therefore, choosing an appropriate learning rate combination is crucial for obtaining good training results.

[0077] like Figure 6 As shown, GS2L converges quickly across different network slices, with improved performance. This demonstrates that the GS2L model can effectively learn and optimize the embedding and resource allocation of security service function chains in various slice environments. Different network slices may have different characteristics and requirements, but the GS2L model can flexibly adapt to these changes and provide efficient services.

[0078] Figure 7 This shows the reward performance of different algorithms across different network slices. For example... Figure 7As shown, the performance of SAC-NSAll is worse than that of SAC-AvgNS1-4, indicating that the SAC model is not suitable for learning slice-specific knowledge from the environment. In contrast, GSAC-NSAll performs similarly to GSAC-AvgNS1-4, mainly because the GSAC model can extract graph information from the network topology, while slice-specific models may encounter performance degradation during training. Ultimately, GS2L-NSAll and GS2L-AvgNS1-4 achieve the best round rewards among all curves and exhibit good stability and learning ability during training. This further demonstrates the effectiveness of the GS2L model in solving the problems of security service function chain embedding and resource allocation.

[0079] from Figure 8 , Figure 9 , Figure 10 In terms of bandwidth and computational resource consumption, the GS2L-NS1 model maintains a moderate level of QoS satisfaction while consuming moderate resources. In contrast, the GS2L-NSAll model achieves similar QoS satisfaction but consumes slightly more bandwidth and computational resources than the GS2L-NS1 model. However, the GSAC-NSAll model, consuming the most resources, only achieves a third-level QoS satisfaction. The PSO and SAC methods perform the worst, with the SAC model using less bandwidth and the PSO method wasting more computational resources.

[0080] In summary, the GS2L model outperforms the GSAC and SAC methods in solving the security service function chain embedding and resource allocation problems. This further validates the effectiveness and importance of Graph Neural Networks (GNNs) and Gated Recurrent Units (GRUs) in the model, providing it with better learning and expressive capabilities, enabling it to better handle high-dimensional problems.

[0081] Specific embodiments of the invention have now been described. Other embodiments are within the scope of the appended claims. In some cases, the actions described in the claims can be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing can be advantageous.

[0082] It should be noted that all directional indicators (such as up, down, left, right, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicator will also change accordingly.

[0083] In the description of this invention, the terms "first" and "second" are used only for convenience in describing different components or names, and should not be construed as indicating or implying a sequential relationship, relative importance, or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" and "second" may explicitly or implicitly include at least one of that feature.

[0084] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0085] It should be noted that although specific embodiments of the present invention have been described in detail with reference to the accompanying drawings, this should not be construed as limiting the scope of protection of the present invention. Various modifications and variations that can be made by those skilled in the art without inventive effort within the scope described in the claims still fall within the scope of protection of the present invention.

[0086] The examples of the embodiments of the present invention are intended to concisely illustrate the technical features of the embodiments of the present invention, so that those skilled in the art can intuitively understand the technical features of the embodiments of the present invention, and are not intended to be an improper limitation of the embodiments of the present invention.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for configuring security services for industrial IoT slicing based on deep reinforcement learning, characterized in that, include: An industrial Internet of Things (IIoT) architecture is established, which includes terminal devices, network slicing modules, virtual secure network functional modules, orchestration modules, and physical networks. In the virtual security network functional module, based on software-defined networking and network function virtualization technology, the user's security service request is modeled to obtain a security service request model. The security service request model includes a virtualized security service function chain, multi-dimensional security levels, and resource pricing factors. A model is trained based on security service network function instances and security service function chains to obtain a security service function chain deployment scheme, wherein: the terminal device sends the security service request to the network operator; the network operator determines the security service network function instance mapped to the corresponding virtual security network service function based on the security service function chain deployment scheme; and the security service function chain deployment scheme is satisfied through a service satisfaction index, wherein the service satisfaction index includes security level and service resources. The security level is calculated by comparing the expected security level with the actual security level based on the security service network function instance and its corresponding virtual security network function. An indicator function is used to indicate whether each correspondence meets the security level. If the expected security level value is greater than or equal to the actual provided security level value, then the correspondence meets the security level. Based on the slice characteristics, request distribution, and simplified action space of network slices, the security service function chain is embedded into the industrial IoT architecture, including: extracting features of the topology and resource allocation of network slices through graph neural networks; obtaining the request distribution of the network slices through GRU model learning and prediction; simplifying the action space through sequence-to-sequence model; and embedding the security service function chain into the industrial IoT architecture based on the slice characteristics, the request distribution, and the simplified action space.

2. The method according to claim 1, characterized in that, In the aforementioned industrial IoT architecture, different network slices have different distributions; The network slicing module adjusts the distribution of the network slices based on the slice features.

3. The method according to claim 1, characterized in that, The resource pricing factors include computing resources and bandwidth resources; The service resources are calculated in the following way: Satisfaction with the computing resources and the bandwidth resources is represented by a utility function.

Images