A Multi-Device Linkage and Control Method for IoT-Based Security Systems
By constructing a unified protocol adaptation layer and a multi-device linkage management and control platform, the problem of interconnection and interoperability of heterogeneous devices in the IoT security system has been solved, achieving efficient and reliable multi-device linkage management and control, improving the system's adaptability and resource utilization, and reducing operation and maintenance costs.
Patent Information
- Application Number
- CN202611319486.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-28
- Publication Date
- 2026-11-17
AI Technical Summary
In existing IoT security systems, heterogeneous devices are difficult to interconnect, multi-device linkage lacks a unified management and control mechanism, and there are timing chaos and logical conflicts in the collaborative response between devices. The system has poor adaptability, low fault tolerance and robustness, high operation and maintenance costs, and is difficult to meet the security needs of complex scenarios.
A unified protocol adaptation layer is constructed to enable heterogeneous device access, a multi-device linkage management and control platform is established, a collaborative response timing management mechanism and strategy conflict resolution are designed, a status monitoring module and dynamic scheduling mechanism are deployed, a security situation assessment and decision optimization system is constructed, and a multi-level fault tolerance mechanism and reinforcement learning algorithm optimization strategy are adopted.
It enables seamless access and full interconnection of heterogeneous devices, ensures the correctness of linkage response timing and the consistency of strategies, improves linkage execution efficiency and reliability, enhances system adaptability and resource utilization, and reduces operation and maintenance costs.
Smart Images

Figure CN123123464A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the technical field of the Internet of Things, and specifically relates to a multi-device linkage management and control method for a security and protection system based on the Internet of Things. Background Art
[0002] With the rapid development of information technology, Internet of Things technology has penetrated into various fields of social production and residents' lives, forming diversified application ecosystems such as smart homes, smart cities, and industrial Internet of Things. In the Internet of Things system architecture, as a key link for ensuring the safety of personnel's lives and property and the stable operation of the system, the security and protection system undertakes core functions such as intrusion detection, environmental monitoring, and early warning disposal, and its importance is increasingly prominent.
[0003] Among them, multi-device linkage management and control, as one of the core technologies of the security and protection system, aims to realize coordinated operation and unified management between security and protection devices with different functions and different protocols. Traditional security and protection systems mostly adopt a mode of independent operation of a single device or simple device combination, which has been difficult to meet the demand for systematic security protection in complex scenarios.
[0004] The existing technology has the following deficiencies: security and protection devices from different manufacturers adopt heterogeneous communication protocols and data formats, making it difficult to achieve interconnection and intercommunication between systems, resulting in a serious information island phenomenon; multi-device linkage lacks a unified management and control mechanism, and there are problems such as timing disorder and logical conflict in the collaborative response between devices, making it difficult to guarantee the efficiency and reliability of linkage execution; the architecture of the security and protection system is rigid, with insufficient capabilities in device status monitoring, fault diagnosis and dynamic scheduling, resulting in poor adaptability when facing complex environments and sudden safety incidents, and a single device fault is likely to cause cascading failure of the linkage chain, leading to low system fault tolerance and robustness. In addition, when the existing technology deals with large-scale Internet of Things security and protection scenarios, the management and control complexity and operation and maintenance costs rise sharply, making it difficult to realize the optimal configuration and efficient utilization of multi-device resources. These technical problems seriously restrict the development of the multi-device linkage management and control capability of the Internet of Things security and protection system, and it is urgent to solve them through technical innovation. Summary of the Invention
[0005] The purpose of the present invention is to provide a multi-device linkage management and control method for a security and protection system based on the Internet of Things, which can effectively solve the problems in the above-mentioned background art.
[0006] To achieve the above purpose, the technical solution adopted by the present invention is as follows: A multi-device linkage control method for a security prevention system based on the Internet of Things, comprising the following specific steps: Step 1: Construct a unified protocol adaptation layer and heterogeneous device access management: design and deploy the unified protocol adaptation layer to realize standardized conversion of communication protocols and data formats of security devices from different manufacturers, complete information collection, capability identification and unified identifier assignment of heterogeneous devices through the device registration module, and seamlessly access heterogeneous security devices to the unified control platform; Step 2: Establish a multi-device linkage control platform and device topology relationship management: build a centralized control platform to realize unified registration, configuration and management of security devices, automatically construct a device topology relationship graph through a device discovery mechanism, determine physical connections and logical associations between devices based on the topology graph, so as to provide basic support for linkage strategy configuration; Step 3: Design a collaborative response timing management mechanism and strategy conflict resolution: establish a timing management module, ensure the timing correctness of multi-device linkage response through a priority scheduling algorithm and a time synchronization mechanism, identify logical conflicts between linkage rules by adopting a strategy conflict detection algorithm, resolve conflicts through a priority arbitration mechanism and generate a conflict-free linkage execution sequence; Step 4: Realize device status monitoring and fault diagnosis: deploy a status monitoring module to collect the operating status, working parameters and performance indicators of each security device in real time, comprehensively analyze the device status by adopting a multi-dimensional health assessment algorithm, identify device faults through an anomaly detection model and trigger a diagnosis process, and generate fault reports and maintenance suggestions; Step 5: Construct a dynamic scheduling and fault tolerance mechanism: establish a dynamic scheduling module to dynamically assign linkage tasks according to the real-time load and health status of devices, realize task load balancing by adopting a weighted round-robin scheduling algorithm, introduce an active-standby switchover mechanism and a task transfer mechanism to ensure the continuity of linkage services, and prevent fault spread through a cascading failure blocking algorithm; Step 6: Construct a security situation assessment and decision optimization system: establish a situation assessment module to comprehensively analyze multi-dimensional data such as device status, alarm frequency and environmental parameters, dynamically assess and grade the overall security situation, and continuously optimize linkage strategy parameters by using a reinforcement learning algorithm based on historical execution data.
[0007] Preferably, the unified protocol adaptation layer in Step 1 supports more than 20 mainstream Internet of Things communication protocols including Modbus, BACnet, ONVIF, OPCUA, MQTT, CoAP, HTTP, HTTPS, WebSocket, DTLS, etc. The protocol adaptation layer is equipped with an extensible protocol registration interface, which supports adding custom protocol drivers in the form of plug-ins to realize fast access of new device types.
[0008] Preferably, the device registration module in Step 1 adopts a two-way authentication mechanism based on digital certificates to ensure the authenticity of device identities. During the device registration process, protocol capability reporting, functional attribute mapping and resource quota allocation are completed. The protocol conversion module adopts a state machine-driven parsing architecture, and realizes asynchronous communication and data buffering through a message queue.
[0009] Preferably, in step 2, the device topology graph is organized using a directed weighted graph structure, where nodes represent security devices, edges represent communication links between devices, and the weights of the edges represent communication latency and reliability assessment values. During the topology construction process, key devices and redundant paths are automatically identified, providing a basis for the reliability design of linkage strategies.
[0010] Preferably, in step 3, the linkage strategy configuration module supports multi-dimensional strategy templates based on device type, security event type, and environmental conditions. The strategy template includes triggering conditions, execution actions, execution order, time constraints, and priority parameters. The strategy engine selects a matching strategy from the strategy library based on the real-time security situation and generates an executable task sequence.
[0011] Preferably, in step 3, the strategy conflict detection algorithm performs static analysis and runtime detection on all strategies in the strategy library. Static analysis is performed during strategy configuration to detect overlapping conditions and contradictory actions. Runtime detection is performed before strategy execution to verify the compatibility of strategies in the current state. Conflict resolution adopts a priority-based arbitration mechanism, and when priorities are the same, the first-come, first-served principle is adopted.
[0012] Preferably, in step 4, the status monitoring module collects device data through the IoT gateway, with a collection period of 2 seconds. The data types include device online status, CPU utilization, memory utilization, number of network connections, business processing volume, and sensor readings. The anomaly detection model adopts an unsupervised learning algorithm based on an autoencoder, and realizes abnormal status recognition through training on normal behavior patterns.
[0013] Preferably, the fault diagnosis process in step 4 includes three stages: fault location, fault classification, and impact assessment. Fault location is determined by Bayesian network inference to identify the faulty equipment and the cause of the fault. Fault classification uses a random forest-based classifier to identify the fault type. Impact assessment analyzes the impact range of the fault on the linkage and generates a fault propagation path diagram.
[0014] Preferably, in step 5, the dynamic scheduling module adopts a task allocation strategy based on real-time load weight. The load weight calculation comprehensively considers four factors: CPU utilization, memory utilization, network bandwidth utilization, and task queue length. The weight calculation formula is a weighted summation form, and the weighting coefficients of each factor are adaptively adjusted according to the device type and task type.
[0015] Preferably, the fault tolerance mechanism in step 5 adopts a redundancy configuration strategy of 1 primary and 2 backups. The critical equipment is deployed with dual primary and backup hot standby. When the primary equipment fails, it automatically switches to the backup equipment. The heartbeat detection period is 500 milliseconds, the fault switching delay is less than 100 milliseconds, and the consistency of the task context is ensured through a state synchronization mechanism during the task transfer process.
[0016] Preferably, the availability assurance of the linkage link in step 5 is achieved through multi-dimensional evaluation. The evaluation indicators include device online rate, alarm accuracy, linkage success rate, response latency and resource utilization. The cascading failure blocking algorithm monitors the device status and link connectivity in real time, and automatically cuts off the fault propagation path and re-plans the linkage link when an anomaly is detected.
[0017] Preferably, in step 6, the security situation assessment adopts a multi-factor fusion algorithm, which comprehensively considers the equipment operating status, alarm event statistics, environmental monitoring data, network traffic characteristics and historical statistical patterns. The situation classification divides the security situation into five levels: normal, attention, alert, threat and severity based on the assessment score. Different levels correspond to different linkage strategy response mechanisms.
[0018] Preferably, in step 6, the strategy optimization adopts a reinforcement learning model based on a deep Q-network. The model takes alarm frequency, equipment load, response latency and situation score as state inputs and linkage rule adjustment as action output. The reward function guides the strategy to evolve in the direction of improving the overall safety performance of the system. The strategy parameter update cycle is 24 hours.
[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. Comprehensive Interoperability: This invention enables heterogeneous device access through a unified protocol adaptation layer, supporting more than 20 mainstream IoT communication protocols and an extensible protocol registration interface. Heterogeneous security devices from different manufacturers and using different protocols can quickly access the system, eliminating information silos and achieving comprehensive interoperability at the device level. This effectively solves the technical problem of difficulty in achieving interoperability between systems in the background technology.
[0020] 2. Highly efficient and reliable linkage control capability: This invention ensures the timing correctness and policy consistency of multi-device linkage response through a timing management module and a policy conflict resolution algorithm. The collaborative response latency is reduced to less than 200 milliseconds, and the automatic policy conflict resolution rate reaches 100%. It effectively solves the problems of timing disorder and logical conflict in the collaborative response between devices in the background technology, and significantly improves the linkage execution efficiency and reliability.
[0021] 3. Intelligent fault tolerance and dynamic scheduling capabilities: This invention achieves continuity assurance for linked services through multi-level fault tolerance mechanisms and dynamic scheduling mechanisms. It adopts a 1-primary-2-backup redundant configuration and task transfer strategy. When the health of the device declines, the task is automatically migrated. The cascading failure blocking algorithm prevents single-point failures from spreading to the entire linkage link. Compared with traditional systems, the single-point failure rate is reduced by more than 90%, and the availability of the linkage link is improved to more than 99.9%, effectively solving the problems of rigid architecture and low fault tolerance in the background technology.
[0022] 4. Adaptive security situation awareness and strategy optimization capability: This invention achieves dynamic situation assessment, five-level hierarchical response, and continuous strategy optimization through a security situation assessment and decision optimization system based on reinforcement learning. The strategy adjustment cycle is shortened from several days of traditional manual configuration to several hours of automatic system optimization. The system's adaptability to complex environments and sudden security events is significantly enhanced, effectively solving the problem of poor adaptability in the background technology.
[0023] 5. Resource optimization and operation and maintenance efficiency improvement capabilities: This invention uses an intelligent scheduling algorithm to comprehensively consider equipment load and health for task allocation, thereby achieving optimized configuration and load balancing of resources across multiple devices. The overall resource utilization rate of the system is increased by more than 40% compared to traditional solutions, and the operation and maintenance cost is reduced by more than 50%, effectively solving the problem of rapidly increasing management complexity and operation and maintenance costs in the background technology. Attached Figure Description
[0024] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall technical solution architecture of the multi-device linkage control method for a security system based on the Internet of Things, according to an embodiment of this application. Figure 2 This is a schematic diagram of the core principle framework of the collaborative response timing management mechanism and strategy conflict resolution in the multi-device linkage control method of the Internet of Things-based security system according to the embodiments of this application; Figure 3 This is a flowchart illustrating the main stages of device topology management and linkage strategy configuration in the multi-device linkage control method of an IoT-based security system according to an embodiment of this application. Figure 4 This is a logical flowchart of the device status monitoring and fault diagnosis method in the multi-device linkage control method of the Internet of Things-based security system according to an embodiment of this application; Figure 5 This is a schematic diagram of the multi-level interaction relationship and data flow of the dynamic scheduling and fault tolerance mechanism in the multi-device linkage control method of the Internet of Things-based security system according to an embodiment of this application. Detailed Implementation
[0025] Example 1 To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.
[0026] The present invention provides a multi-device linkage control method for a security prevention system based on the Internet of Things. The method includes six core steps, which are respectively constructing a unified protocol adaptation layer and heterogeneous device access management, establishing a multi-device linkage control platform and device topology relationship management, designing a collaborative response timing management mechanism and policy conflict resolution, realizing device status monitoring and fault diagnosis, constructing a dynamic scheduling and fault tolerance mechanism, and constructing a security situation assessment and decision optimization system. The specific implementation manner of each step is described in detail below.
[0027] Step 1: Construct a unified protocol adaptation layer and heterogeneous device access management Design and deploy a unified protocol adaptation layer to realize standardized conversion of communication protocols and data formats of security devices from different manufacturers. As the core communication hub between heterogeneous devices and the control platform, the unified protocol adaptation layer undertakes key functions such as protocol parsing, data conversion, and message routing. The internal architecture of the protocol adaptation layer adopts a layered design, which sequentially includes a physical access layer, a protocol parsing layer, a data conversion layer, and a standardized output layer from bottom to top. The physical access layer is responsible for establishing underlying communication connections with various security devices, supporting multiple physical access methods such as serial port, network, and wireless; the protocol parsing layer performs in-depth parsing for each accessed communication protocol, and extracts original data and control instructions reported by the devices; the data conversion layer converts the parsed heterogeneous data formats into a unified standardized data model inside the platform; the standardized output layer provides services to upper-layer applications in the form of a unified interface for the converted data.
[0028] The unified protocol adaptation layer supports more than 20 mainstream Internet of Things communication protocols including Modbus, BACnet, ONVIF, OPC UA, MQTT, CoAP, HTTP, HTTPS, WebSocket, DTLS, etc., covering common communication scenarios in the security field such as industrial control, building automation, video surveillance, and device interconnection. For each supported protocol type, the protocol adaptation layer has a built-in corresponding protocol driver module, and each driver module includes four sub-components: a protocol codec, a connection manager, a heartbeat maintainer, and an exception handler. The protocol codec is responsible for parsing protocol messages in binary or text format into structured data objects, or encapsulating structured instructions into message formats that conform to protocol specifications; the connection manager is responsible for maintaining communication sessions with devices, and processing life cycle events such as connection establishment, maintenance, and disconnection; the heartbeat maintainer is responsible for sending heartbeat detection messages according to the protocol specifications and detecting the response status of the peer end to ensure the activity of the communication link; the exception handler is responsible for capturing various abnormal situations during the protocol interaction process, and triggering corresponding recovery or alarm mechanisms.
[0029] The protocol adaptation layer is equipped with an extensible protocol registration interface, supporting the addition of custom protocol drivers via plugins. Protocol plugins employ a dynamic loading mechanism; each plugin package exists as an independent deployment unit, containing protocol driver code, configuration metadata, and resource files. Upon platform startup, it scans the preset plugin directory, automatically identifying and loading valid protocol plugins. During runtime, hot-swapping is supported, allowing new protocol types to be dynamically registered without affecting existing device communication. When integrating security equipment from new vendors, developers only need to write driver code according to the protocol plugin development specifications and package and deploy it to quickly achieve integration of new device types without modifying the platform's core framework.
[0030] The device registration module completes the information collection, capability identification, and unified identifier allocation for heterogeneous devices. As the entry point to the device access management platform, the device registration module is responsible for executing the entire process of device authentication, information registration, capability reporting, and identifier allocation. Internally, the device registration module comprises five core components: a registration controller, an authentication engine, an information parser, an identifier allocator, and a registration storage. The registration controller coordinates the execution order and state transitions of the entire registration process, ensuring that each component completes its task sequentially. The authentication engine verifies the authenticity of the identity credentials submitted by the device. The information parser extracts key information such as device model, firmware version, and feature list from the device registration request. The identifier allocator assigns a globally unique device identifier to each access device. The registration storage persists the registered device information to the device registration database.
[0031] The device registration module employs a two-way authentication mechanism based on digital certificates to ensure the authenticity of device identities. The device side has a pre-installed device certificate issued by the platform, containing a unique device identifier, public key information, and CA signature. The platform side configures a corresponding CA root certificate to verify the legitimacy of the device certificate. During device registration, a certificate exchange and verification process is first performed. The device submits its certificate to the platform, which verifies the certificate signature, validity period, and revocation status. After confirming the certificate's validity, the platform sends its own certificate to the device. The device then verifies the legitimacy of the platform certificate, completing the two-way authentication. After successful certificate verification, a TLS-based encrypted communication channel is established between the device and the platform. All subsequent registration data exchanges take place within this encrypted channel. This two-way authentication mechanism effectively prevents unauthorized device access and man-in-the-middle attacks, ensuring the security of device access.
[0032] During device registration, protocol capability reporting, functional attribute mapping, and resource quota allocation are completed. After completing identity authentication, the device sends a capability reporting message to the platform. The message contains capability description information such as a list of supported communication protocols, a list of functional modules, sensor configurations, and alarm types. The platform receives and parses the capability reporting message, calling the functional attribute mapper to convert the device-specific capability description into a unified capability model representation within the platform, establishing a mapping relationship between device capabilities and platform functional interfaces. The resource quota allocator allocates system resource quotas to the device based on device type and access policy, including maximum connection limits, message sending frequency limits, and data cache space quotas. Resource quota information is persistently stored along with the device identifier for subsequent access control and traffic management.
[0033] The protocol conversion module employs a state machine-driven parsing architecture to achieve efficient and reliable protocol data processing. The state machine parser maintains a dedicated parsing state machine for each protocol type, defining the valid parsing paths and state transition rules for protocol messages. During parsing, the state machine transitions based on the current state and input characters, synchronously updating the parsing context and accumulating parsing results. When encountering illegal input or state transition errors, the state machine enters an error handling branch, triggering a data verification failure alarm and discarding the current message. The state machine-driven parsing architecture offers advantages such as high parsing efficiency, strong error detection capabilities, and clear code structure, making it suitable for processing various structured and semi-structured protocol data.
[0034] The protocol conversion module implements asynchronous communication and data buffering through a message queue. The protocol parsing layer and the data conversion layer are decoupled via the message queue. The protocol parsing submodule encapsulates the parsed raw data into data messages and publishes them to the message queue. The data conversion submodule subscribes to and consumes data messages from the message queue for conversion processing. The message queue acts as a data buffer, smoothing out differences in processing rates and avoiding data loss or blocking caused by a mismatch between data generation and consumption rates. The message queue employs a persistent storage strategy; unconsumed messages are temporarily stored on disk, allowing processing to resume after system recovery, ensuring data processing reliability. The message queue also supports message priority settings, allowing high-priority alarm data to be consumed and processed first, reducing end-to-end processing latency for critical information.
[0035] The unified protocol adaptation layer seamlessly connects heterogeneous security devices to the unified management and control platform. After completing the registration process, the devices enter a normal operating state. The data generated by the devices undergoes full-process processing by the protocol adaptation layer, and is finally output to the upper-layer application module in a standardized format; the control instructions issued by the platform undergo reverse processing by the protocol adaptation layer, are converted into a protocol message format recognizable by the devices, and then issued to the target devices. The entire data interaction process is transparent to upper-layer applications and underlying devices. The application layer does not need to care about the specific communication protocol details of the devices, and the device side does not need to understand the data format specifications of the platform, achieving true cross-vendor and cross-protocol seamless integration.
[0036] Step 2: Establish the topology relationship management between the multi-device linkage management and control platform and the devices Build a centralized management and control platform to realize unified registration, configuration and management of security devices. As the core control hub of the multi-device linkage management and control system, the centralized management and control platform undertakes core management functions such as device resource management, linkage strategy management, task scheduling management, and system configuration management. The platform adopts a microservice architecture design, splitting each management function into independent service modules, including device management service, strategy management service, task scheduling service, configuration management service, authentication and authorization service, log audit service, etc. Service discovery and load balancing are realized between each service module through the service registration center, asynchronous communication and event distribution are realized through the message bus, and fast access to shared data is realized through the distributed cache. The microservice architecture endows the platform with good scalability and high availability. When the system load increases, the number of instances of the bottleneck service can be expanded in a targeted manner. When a certain service instance fails, it can automatically switch to a healthy instance to ensure the overall availability of the system.
[0037] The device management service is responsible for maintaining the add, delete, modify and query operations of device registration information, and supports functions such as batch device import and export, device group management, and device tag management. Device group management allows users to logically group devices according to dimensions such as physical location, functional area, and device type, which facilitates batch operations and permission control; device tag management supports adding custom business tags to devices, and the tag information can be used for strategy matching conditions and statistical report dimensions. The strategy management service is responsible for managing the full life cycle operations of linkage strategies, including configuration, enabling, disabling, and version control, and supports strategy template management and strategy import and export functions. The task scheduling service is responsible for receiving linkage trigger events and generating corresponding execution tasks, distributing the tasks to target devices or execution nodes according to the scheduling algorithm, monitoring the task execution status and handling abnormal situations.
[0038] The device discovery mechanism automatically constructs a device topology map. It employs a hybrid discovery strategy combining active probing and passive monitoring. In active probing mode, the management platform periodically sends probe packets to devices within the network segment and determines their online status based on the responses. In passive monitoring mode, the management platform listens for device status change notifications in the network, receiving and processing notification messages when devices go offline, offline, or undergo configuration changes. This hybrid discovery strategy balances timeliness and resource efficiency, enabling timely capture of device status changes while avoiding excessive network probing overhead.
[0039] The device discovery module collects basic information such as device network address, port number, MAC address, manufacturer identifier, device model, firmware version, and operating status, as well as network topology information such as device communication port mapping table, gateway device association, and aggregation switch cascading relationships. After the device information is collected, it is sent to the topology building engine for topology relationship analysis. The topology building engine first calculates the network segment to which the device belongs based on the device IP address and subnet mask, and establishes the network layer connection relationship between devices; then it analyzes the link layer adjacency relationship between devices based on MAC address learning records and ARP entries; finally, it integrates network layer and link layer information, supplemented by topology-related data in the device's SNMP management information database, to construct a complete device network topology map.
[0040] The device topology graph is organized using a directed weighted graph structure. Nodes in the directed weighted graph represent security devices, with each type of device corresponding to a specific type of node object. Node objects contain attribute fields such as device identifier, device type, capability attributes, and status information. Edges in the directed weighted graph represent communication links between devices, with each edge corresponding to a physical or logical communication path. Edge objects contain attribute fields such as source node identifier, destination node identifier, link type, bandwidth capacity, and transmission medium. The edge weights represent communication latency and reliability assessment values. Weight calculation comprehensively considers multiple factors such as link transmission latency, routing hop count, link bandwidth utilization, and historical packet loss rate, obtaining a comprehensive evaluation score through a weighted summation method. The topology graph supports dynamic updates; when device status changes or link attributes change, the graph data is updated synchronously, triggering an association analysis algorithm to reassess the rationality of the topology structure.
[0041] During topology construction, critical devices and redundant paths are automatically identified. The critical device identification algorithm performs centrality analysis on the topology graph, calculating the degree centrality, betweenness centrality, and proximity centrality of each device node. Devices with high centrality scores have a significant impact on network connectivity and linkage functions and are marked as critical devices with higher priority monitoring strategies. The redundant path identification algorithm performs loop detection and minimum cut analysis on the topology graph, identifying redundant paths available for failover, providing fundamental data support for ensuring the reliability of linkage links. When a critical device fails, the system can quickly locate redundant paths and perform link switching, maintaining the continuity of linkage services.
[0042] The physical connections and logical relationships between devices are determined based on the topology map. Physical connection analysis calculates the physically reachable paths between devices based on device network coordinates and link attributes, generating a physical connection topology map. Logical relationship analysis constructs a logical dependency graph between devices based on device service functions and data flow. The physical connection graph and the logical dependency graph together constitute a complete device relationship model, providing multi-dimensional relationship information support for configuring linkage strategies. When configuring cross-device linkage strategies, the system automatically verifies whether the relationship paths between all devices involved in the strategy are unobstructed. If a path is broken, an alarm is issued, guiding the user to adjust the strategy configuration or repair the physical connection.
[0043] The device topology map provides the foundation for configuring linkage strategies. When creating or modifying a linkage strategy, the strategy configuration module calls the topology query interface to obtain the topology information of the devices involved in the strategy, including physical connection paths between devices, logical dependencies, and communication latency assessment values. Topology information serves as one of the input conditions for strategy verification. The system verifies the rationality of the strategy configuration, such as verifying whether the response capability of the action-executing device meets timing requirements and whether a reachable path exists between the target device and the triggering device. When the strategy is issued and executed, the scheduling module plans the task routing path based on the topology information, selecting the optimal or alternative execution route to ensure that the linkage command can reliably reach the target device.
[0044] Step 3: Design a collaborative response timing management mechanism and strategy conflict resolution. Establish a timing management module, which ensures the timing correctness of multi-device linkage response through a priority scheduling algorithm and a time synchronization mechanism. The timing management module is a core component that guarantees the correct execution of coordinated actions of multiple devices, and is responsible for maintaining the time reference of all linkage tasks in the system, coordinating the execution sequence of concurrent tasks, and controlling the precise timing of action triggering. The timing management module internally includes four sub-modules: a time reference service, a scheduling engine, a timing controller and a synchronization manager. The time reference service maintains a unified time reference source of the system, supports two time synchronization modes: NTP Network Time Protocol synchronization and GPS clock synchronization, and ensures the time consistency of all nodes in the system; the scheduling engine performs task sorting and scheduling decision-making according to task priority and resource availability; the timing controller generates an accurate execution schedule according to the timing constraints configured by the policy, and controls the start moment of each execution node; the synchronization manager coordinates the action synchronization between distributed execution nodes, ensuring that all nodes execute specified actions simultaneously at a specified moment.
[0045] The priority scheduling algorithm adopts a multi-level feedback queue scheduling strategy, assigns the to-be-executed linkage tasks to multiple priority queues, and tasks in the high-priority queue obtain scheduling opportunities preferentially. Tasks of the same priority are scheduled in a round-robin manner with time slices. Each task yields the processor to other tasks in the same queue after executing one time slice, preventing low-priority tasks from being starved indefinitely. The task priority is dynamically calculated according to factors such as the priority parameter configured by the policy, the emergency degree of events, and the device load status. Tasks corresponding to unexpected events and emergency alarms are assigned higher priorities to ensure that critical responses can be executed quickly. The scheduling algorithm also considers the time constraint relationship between tasks. Tasks with sequential constraints are assigned to the same scheduling batch, and subsequent tasks can be scheduled only after the pre-task is completed, so as to ensure the orderliness of task execution.
[0046] The time synchronization mechanism uses the Precision Time Protocol to achieve microsecond-level time synchronization accuracy. All execution nodes in the system run a time synchronization client, which maintains a continuous time synchronization relationship with the time reference server. The time synchronization client periodically sends a synchronization request message to the time server, and records the local timestamp of the message sending moment; after receiving the server response message, it records the local timestamp of the receiving moment, calculates the time offset according to the round-trip delay and corrects the local clock. The stability of time synchronization accuracy is improved by taking the average value of multiple measurements, and the synchronization period is set to 1 second, ensuring that the time deviation of each node in the system at any moment is controlled within an acceptable range.
[0047] A strategy conflict detection algorithm is employed to identify logical conflicts between linked rules. This algorithm performs two phases of conflict resolution: static analysis and runtime detection, examining all strategies in the strategy library. Static analysis is performed during strategy configuration. When a user submits a new strategy configuration or modifies an existing strategy, the system immediately performs a comprehensive scan of the strategy library to detect any overlapping conditions or contradictory actions between the new and existing strategies. Overlapping conditions refer to the intersection of triggering conditions between two strategies, potentially leading to the simultaneous triggering of multiple strategies for the same event. Contradictory actions refer to conflicting actions performed by two strategies on the same device or controlled object; for example, one strategy requires the device to be turned on while another requires it to be turned off. The static analysis phase uses set theory and Boolean algebra methods to formally model the strategy conditions, transforming the overlapping condition detection into a set intersection operation problem.
[0048] Runtime checks are performed before policy execution to verify policy compatibility in the current state. When a linked event triggers policy execution, the timing management module queries the currently active policy set before scheduling the task to check for conflicts between the policy to be executed and the already active policies. Runtime checks need to consider the timeliness and state dependencies of policies; the same policy may be in different states at different times. For example, some policies may have time-limited validity periods, meaning they are only effective within a specific time frame. The runtime check algorithm tracks the execution status of each policy by establishing a policy activation status table, and makes a comprehensive judgment based on the activation status table and policy time constraints during checks.
[0049] Policy conflict resolution employs a priority-based arbitration mechanism. When a policy conflict is detected, the system arbitrates based on the priority parameters of the conflicting policies. The higher-priority policy is granted execution rights, while the lower-priority policy is temporarily suspended or rejected. The priority parameter is specified by the user during policy configuration; a higher value indicates a higher priority. Priority arbitration follows a deterministic principle. Policy conflicts of equal priority are handled on a first-come, first-served basis: the policy triggered first is executed first, and the policy triggered later is rejected, with the conflict log recorded for subsequent handling by operations personnel. The conflict arbitration result is notified to the relevant policy configuration parties, allowing users to understand the policy execution status and make necessary configuration adjustments.
[0050] A priority arbitration mechanism resolves conflicts and generates conflict-free linked execution sequences. The linked execution sequence generation algorithm takes the policy trigger event sequence as input, comprehensively considering factors such as policy priority, time constraints, and resource dependencies, and outputs an ordered task execution plan. During sequence generation, trigger events are first sorted according to their occurrence time, and then the policy corresponding to each event is processed sequentially. For conflicting policies, an arbitration decision is made; policies that fail arbitration are not added to the execution sequence. The generated execution sequence is then sent to the scheduling engine in the form of task scheduling, and the scheduling engine schedules and executes each linked task according to the sequence order and time constraints.
[0051] The linkage policy configuration module supports multi-dimensional policy templates based on device type, security event type, and environmental conditions. A policy template defines the basic framework of a policy, including five core elements: trigger conditions, execution actions, execution order, time constraints, and priority parameters. Trigger conditions define the event conditions for a policy to take effect, supporting combinations of various condition types such as device status change conditions, alarm event conditions, time conditions, and environmental parameter threshold conditions. Execution actions define the sequence of actions to be executed after the policy is triggered, including device control actions, alarm notification actions, log recording actions, and policy linkage actions. Execution order defines the sequential relationship between multiple actions. Time constraints define the effective time period and execution time requirements of the policy. Priority parameters define the priority value of the policy during conflict arbitration. The policy engine selects matching policies from the policy library based on the real-time security situation and generates an executable task sequence.
[0052] The strategy engine's workflow comprises four stages: situation awareness, strategy matching, sequence generation, and task deployment. The situation awareness stage collects multi-source information such as current system status, active alarms, and environmental data to construct a feature vector of the current situation. The strategy matching stage matches the situation feature vector with the trigger conditions of each strategy in the strategy library, selecting a set of strategies that meet the trigger conditions. The sequence generation stage performs conflict detection and arbitration on the successfully matched strategies, generating a conflict-free execution sequence. The task deployment stage converts the execution sequence into specific task scheduling instructions and distributes them to the corresponding execution nodes. The strategy engine continuously loops through the entire process from situation awareness to task deployment, enabling real-time response to the security situation.
[0053] The load weight is calculated using a weighted summation formula, as shown below:
[0054] in, Indicates equipment The overall load weight, Indicates equipment CPU utilization normalized value, Indicates equipment The normalized value of memory usage. Indicates equipment The normalized value of network bandwidth utilization. Indicates equipment Normalized value of task queue length , , , These are the weighting coefficients for the four factors mentioned above. The weighting coefficients are adaptively adjusted based on device type and task type; for computationally intensive tasks, the weight of CPU utilization is appropriately increased, and for data transmission-intensive tasks, the weight of network bandwidth utilization is appropriately increased. The weighting coefficients are obtained through training on historical operating data, and the coefficient values are optimized using a gradient descent algorithm to minimize the deviation between the load weight assessment results and the actual load situation.
[0055] Step 4: Implement equipment status monitoring and fault diagnosis A status monitoring module is deployed to collect real-time operational status, working parameters, and performance indicators of various security devices. As the sensory organ for the system's healthy operation, the status monitoring module comprehensively grasps the device's operating status through multi-dimensional data collection methods. The status monitoring module comprises four core components: a data collector, a time-series database, a real-time computing engine, and an alarm manager. The data collector is responsible for acquiring status data reported by devices from the IoT gateway, supporting both polling and event reporting modes. The time-series database stores the collected time-series status data, supporting efficient time-range querying and aggregation calculations. The real-time computing engine performs online analysis of the time-series data, detects abnormal states, and calculates health indicators. The alarm manager generates alarm events based on the anomaly detection results and triggers corresponding processing flows.
[0056] The status monitoring module collects device data through an IoT gateway. The IoT gateway, deployed at the device access layer, acts as a data aggregation node between devices and the management platform, responsible for protocol conversion, data cleaning, edge computing, and data forwarding. The data collection cycle is set to 2 seconds, meaning that all online devices are polled every 2 seconds to obtain the latest status snapshot. Polling requests are sent to the target devices through a unified protocol adaptation layer. The devices respond to the requests and report their current status data. This data is then aggregated and processed by the gateway before being forwarded to the status monitoring module for analysis. For devices that support proactive reporting, the device proactively pushes a status change notification to the gateway when its status changes significantly. Upon receiving the notification, the gateway immediately forwards it to the status monitoring module, enabling real-time perception of status changes.
[0057] The types of data collected include device online status, CPU utilization, memory utilization, network connections, transaction volume, and sensor readings. Device online status reflects whether the device is currently online and communicable, determined through a heartbeat detection mechanism. CPU utilization reflects the usage of the device's computing resources; memory utilization reflects the usage of the device's storage resources; network connections reflect the number of communication connections currently established by the device; transaction volume reflects the number of business requests processed by the device per unit of time; and sensor readings reflect physical quantity data collected by various sensors attached to the device, such as temperature, humidity, voltage, and current. Different types of security equipment use different combinations of monitoring indicators based on their functional characteristics. For example, video surveillance equipment focuses on monitoring CPU utilization and video stream bitrate, while industrial control equipment focuses on monitoring sensor readings and communication connection status.
[0058] A multi-dimensional health assessment algorithm is employed to comprehensively analyze equipment status. This algorithm establishes an equipment health assessment model, mapping collected multi-dimensional status data into a comprehensive health score. The health assessment model is designed using the analytic hierarchy process (AHP), with a structure comprising three layers: an indicator layer, a criterion layer, and a target layer. The indicator layer corresponds to various raw monitoring indicators; the criterion layer categorizes indicators into three evaluation criteria: performance, reliability, and safety; and the target layer outputs the comprehensive health score. The score for each indicator is calculated using different scoring functions based on the indicator type and value range. Performance and reliability indicators use negative scoring functions, with higher values resulting in lower scores; safety indicators use positive scoring functions, with higher values resulting in higher scores. The criterion layer aggregates scores from subordinate indicators using a weighted average method, with weights determined by expert evaluation based on indicator importance. The target layer aggregates the scores from the three criteria using a linear weighting method to obtain the final comprehensive health score. The comprehensive health score ranges from 0 to 100, with higher scores indicating better equipment health.
[0059] An anomaly detection model identifies equipment faults and triggers diagnostic procedures. The anomaly detection model employs an unsupervised learning algorithm based on an autoencoder, trained using normal behavior patterns to identify abnormal states. The autoencoder consists of an encoder and a decoder. The encoder maps the input state data to a low-dimensional latent space, and the decoder reconstructs the original input data from the latent space representation. During model training, only normal state data samples are used, with the training objective being to minimize the reconstruction error of normal samples. After training, the reconstruction error of normal samples is relatively small, while the reconstruction error of abnormal samples increases significantly due to distribution differences. During anomaly detection, real-time collected state data is input into the trained autoencoder model, the reconstruction error is calculated, and compared with a preset anomaly threshold. Samples with errors exceeding the threshold are identified as abnormal states, triggering the fault diagnosis process.
[0060] The fault diagnosis process includes three stages: fault location, fault classification, and impact assessment. In the fault location stage, Bayesian network inference is used to determine the faulty equipment and its cause. The Bayesian network uses equipment components as nodes and causal relationships between components as edges to establish a fault propagation model; the network's conditional probability table records the prior and conditional probabilities of each component's failure. When a fault occurs, based on observed abnormal symptoms, the posterior probability of each component's failure is calculated using Bayesian network inference; the component with the highest posterior probability is identified as the most likely fault point. Bayesian network inference employs either exact or approximate inference algorithms, with the appropriate inference method selected based on the network size.
[0061] In the fault classification phase, a random forest-based classifier is used to identify fault types. The random forest consists of multiple decision trees, each trained on a randomly selected subset of features and samples. The final classification result is obtained by summing the votes from all decision trees. The fault classifier takes device state features as input and outputs a fault type label. Fault types include hardware faults, software faults, communication faults, configuration errors, and resource exhaustion. The classifier is trained using historical fault case data, and hyperparameter configuration is optimized through cross-validation to ensure the generalization performance of the classification model.
[0062] Impact assessment analyzes the scope of a fault's impact on linked links and generates a fault propagation path diagram. The impact assessment algorithm constructs a fault impact model based on the device topology graph and linkage strategy configuration data. The algorithm first locates the faulty device in the topology graph and identifies all linked links with that device as an upstream node; then, it recursively traverses along the link directions, recording each affected link and its downstream devices; finally, it generates a fault propagation path diagram, visually displaying the scope of the fault's impact. The impact assessment results guide fault handling decisions. When the fault's impact scope is large, the system recommends isolation measures; when redundant paths exist, the system automatically triggers link switching.
[0063] The fault diagnosis process generates fault reports and maintenance recommendations. The fault report includes the fault occurrence time, faulty equipment information, fault type, fault cause analysis, and impact assessment, and is output in a structured document format for maintenance personnel to review. Maintenance recommendations are generated based on the fault type and historical maintenance records, including recommended handling measures, required maintenance tools, and estimated processing time, guiding maintenance personnel to quickly complete fault repairs.
[0064] Step 5: Construct a dynamic scheduling and fault tolerance mechanism A dynamic scheduling module is established to dynamically allocate collaborative tasks based on the real-time load and health status of devices. This module serves as the central scheduling hub for multi-device coordination, responsible for efficiently and reliably allocating collaborative tasks to appropriate execution nodes. The dynamic scheduling module comprises four core components: a task receiver, a scheduling decision-maker, a task dispatcher, and a status tracker. The task receiver receives task requests from the collaborative strategy engine, performs validity checks and parameter parsing, generates internal task objects, and adds them to the task waiting queue. The scheduling decision-maker retrieves tasks from the waiting queue and selects the optimal target execution node based on the scheduling algorithm and device status. The task dispatcher sends the task to the selected execution node and establishes a task execution context. The status tracker tracks task execution progress and handles various execution results such as task completion, timeout, and failure.
[0065] A weighted round-robin scheduling algorithm is used to achieve task load balancing. This algorithm introduces a weighting factor into the traditional round-robin scheduling algorithm, with the weight reflecting the relative processing capacity of the nodes. During scheduling, the system maintains a weighted round-robin queue. Nodes in the queue are allocated scheduling time slices according to their weight ratios, with nodes having stronger processing capabilities receiving more time slices and scheduling opportunities. Device weights are dynamically adjusted based on real-time load status; devices with lower loads receive higher scheduling weights, and devices with higher loads receive lower scheduling weights. During scheduling, the scheduling decision-maker traverses the weighted round-robin queue, assigning tasks to each node sequentially and updating the node load statistics. After assigning several tasks consecutively, the node weights are recalculated and the queue order is adjusted to achieve dynamic load balancing.
[0066] Load weight calculation comprehensively considers four factors: device CPU utilization, memory utilization, network bandwidth utilization, and task queue length. As mentioned earlier, the load weight calculation formula is:
[0067] in , , , The four weighting coefficients are adaptively adjusted based on device type and task type. For general-purpose computing servers, CPU utilization and memory usage have higher weights; for network communication devices, network bandwidth utilization has a higher weight; and for devices with severe task backlog, task queue length has a higher weight. This adaptive adjustment of the weighting coefficients is achieved through a reinforcement learning algorithm, which optimizes the coefficient values based on task completion rate and resource utilization efficiency.
[0068] A primary / backup failover mechanism and a task transfer mechanism are introduced to ensure the continuity of linked services. The fault tolerance mechanism adopts a redundant configuration strategy of 1 primary and 2 backups. Critical equipment is deployed in a dual-machine hot standby architecture, with one primary device and two backup devices forming a high-availability unit. Under normal operating conditions, the primary device carries the entire service load, while the backup devices are in hot standby standby mode, synchronizing the status data of the primary device in real time. When the primary device fails, the system automatically switches services to the backup device to continue providing services. The backup device failover uses a hot standby method, and the failover process is transparent to upper-layer applications; service connections do not need to be rebuilt. The task transfer mechanism plays a role during failover. When the primary device is unavailable, the system automatically transfers tasks currently executing and pending on the primary device to the backup device for continued processing, ensuring that no tasks are lost and execution is uninterrupted.
[0069] Key parameters of the primary / standby switchover mechanism include the heartbeat detection period and the failover latency. The heartbeat detection period is set to 500 milliseconds, meaning the system sends a heartbeat probe message to the primary device every 500 milliseconds, using the heartbeat response to determine the primary device's viability. The heartbeat detection employs a two-way confirmation mechanism; the primary device responds with an acknowledgment response upon receiving the probe message. Response timeouts or errors are considered heartbeat failures. After several consecutive heartbeat failures, the system determines that the primary device has failed, triggering the failover process. The failover latency is defined as the time interval from failure confirmation to the standby device taking over services. The failover latency is required to be less than 100 milliseconds to ensure the impact of the switchover process on services remains within acceptable limits. Controlling the failover latency largely depends on the efficiency of the state synchronization mechanism.
[0070] During task transfer, a state synchronization mechanism ensures the consistency of the task context. This mechanism establishes a real-time data synchronization channel between the primary and backup devices, ensuring that state changes generated during task execution on the primary device are synchronized to the backup device in real time. Synchronized content includes all task context-related information such as task execution progress, intermediate calculation results, configuration parameter changes, and cached data updates. State synchronization employs asynchronous replication; the primary device returns a response immediately after completing a state update, and the background asynchronously synchronizes the update to the backup device. The synchronization channel features message persistence and retransmission mechanisms to ensure that state data is not lost in the event of network jitter or brief interruptions. When a failover occurs, the backup device restores the task context based on the synchronized state data and continues executing any unfinished tasks.
[0071] Cascading failure prevention algorithms prevent fault propagation. Cascading failure refers to a chain reaction where a failure in one device or component in a system causes the impact of the failure to spread along related paths, leading to failures in more devices. Cascading failure prevention algorithms monitor device status and link connectivity in real time, automatically implementing isolation measures to cut off the fault propagation path when anomalies are detected. The algorithm first constructs a fault propagation model based on device topology and inter-device dependencies, describing the propagation path and scope of the fault within the system. The real-time monitoring module continuously collects device status and link metrics data. When an abnormal device status or decreased link connectivity is detected, it immediately analyzes the possible propagation direction of the fault. The algorithm predicts the set of affected downstream devices based on the fault propagation model, calculates the isolation point location, and generates isolation commands. These commands are sent to the relevant devices to perform port closure or link interruption operations. After isolation, the system reassesses the connectivity and availability of the remaining devices. If redundant paths exist, link switching is automatically triggered to restore the inter-device services affected by the isolation.
[0072] The availability of the linkage system is ensured through multi-dimensional evaluation. The evaluation metrics include five core indicators: device online rate, alarm accuracy rate, linkage success rate, response latency, and resource utilization. Device online rate is the ratio of device online time to total online time within a given period, reflecting device availability. Alarm accuracy rate is the ratio of valid alarms to total alarms, reflecting system alarm quality. Linkage success rate is the ratio of successfully completed linkage tasks to total tasks, reflecting linkage execution reliability. Response latency is the time interval from triggering to completion of a linkage task, reflecting system response speed. Resource utilization rate is the degree of device resource occupancy, reflecting resource allocation efficiency. The system continuously collects and statistically analyzes the above metrics. When a metric shows a deteriorating trend, an early warning is triggered to alert maintenance personnel for attention and handling.
[0073] The cascading failure prevention algorithm monitors device status and link connectivity in real time. Upon detecting an anomaly, it automatically cuts off the fault propagation path and re-plans the linked links. After the fault path cutting operation is executed, the original linked links may malfunction due to path breakage. The system automatically initiates the linked link replanning process, searching for alternative paths based on the updated device topology. The replanning algorithm uses the original link start and end points as constraints, maximizes path reliability as the optimization objective, and employs Dijkstra's shortest path algorithm or its variants to search for the optimal path satisfying the constraints in the topology. After finding an alternative path, the system updates the link routing table and distributes it to the relevant devices, allowing linked services to resume execution on the new path. The entire fault handling process is transparent to upper-layer applications, ensuring high availability of linked services.
[0074] Step 6: Construct a security situation assessment and decision optimization system A situation assessment module is established to comprehensively analyze multi-dimensional data such as equipment status, alarm frequency, and environmental parameters to dynamically assess and classify the overall security situation. As the system's intelligent decision-making center, the situation assessment module is responsible for fusing and analyzing massive amounts of multi-source data to form a quantitative assessment conclusion of the overall security status. The situation assessment module comprises four core components: a data fusion unit, an assessment engine, a classifier, and a visualization component. The data fusion unit receives input data from multiple data sources, including the equipment status monitoring module, alarm management module, environmental monitoring module, and network traffic analysis module, and performs data cleaning, format unification, and time alignment. The assessment engine executes a situation assessment algorithm based on the fused data, calculating assessment scores for each dimension and a comprehensive assessment score. The classifier determines the situation level based on the assessment scores. The visualization component displays the assessment results to the user in chart form.
[0075] The security situation assessment employs a multi-factor fusion algorithm. The algorithm design comprehensively considers five dimensions of assessment factors: equipment operating status, alarm event statistics, environmental monitoring data, network traffic characteristics, and historical statistical patterns. The equipment operating status factor originates from the status monitoring module, extracting the health score of each device as input. The alarm event statistics factor comes from the alarm management module, statistically analyzing the number of alarms per unit time, the distribution of alarm types, and the timeliness of alarm handling. The environmental monitoring data factor comes from the environmental sensor network, collecting environmental parameters such as temperature, humidity, smoke concentration, and water immersion status. The network traffic characteristics factor comes from the network traffic analysis module, statistically analyzing network indicators such as traffic volume, protocol distribution, and changes in connection count. The historical statistical pattern factor comes from a historical database, providing historical baselines and fluctuation ranges for each assessment indicator.
[0076] The multi-factor fusion algorithm employs a weighted fusion model to calculate the comprehensive situation score. The algorithm first standardizes the raw data of the five evaluation factors, mapping index data with different dimensions and value ranges to a unified score range of 0 to 100. Then, it assigns weight coefficients based on the degree of influence of each factor on the security situation. Finally, it calculates the comprehensive situation score through a weighted summation method. The weight coefficients are determined using the analytic hierarchy process (AHP), inviting security experts to conduct pairwise comparisons of each factor and construct a judgment matrix. The weight vector of each factor is calculated using the matrix eigenvalues. A higher comprehensive situation score indicates a better security situation, while a lower score indicates a more severe security situation.
[0077] The security situation is divided into five levels: normal, attention, alert, threat and severe based on the evaluation score of the situation grading basis. The normal level corresponds to a comprehensive score of 90 to 100, indicating that the system operates normally without obvious security risks; the attention level corresponds to a score of 75 to 89, indicating that there are potential risk factors that require attention but no action is needed for the time being; the alert level corresponds to a score of 60 to 74, indicating that clear signs of security threats have emerged and preventive measures need to be taken; the threat level corresponds to a score of 45 to 59, indicating that security threats have occurred or are about to occur, and disposal measures need to be taken immediately; the severe level corresponds to a score of 0 to 44, indicating that the security situation is severe, the system may suffer major damage, and an emergency response mechanism must be activated. Different levels correspond to differentiated linkage strategy response mechanisms, and the higher the level, the stricter the triggered linkage strategy and the more aggressive the response actions.
[0078] The reinforcement learning algorithm is used to continuously optimize the parameters of the linkage strategy based on historical execution data. The strategy optimization module adopts a reinforcement learning model based on a Deep Q-Network, and learns the optimal strategy parameter configuration through continuous interaction with the environment. The Deep Q-Network model takes the alarm frequency, device load, response delay and situation score as state inputs, and takes the adjustment of linkage rules as action outputs. The action space is defined as a set of strategy parameter adjustment operations, including trigger threshold adjustment, execution action addition and subtraction, priority adjustment, timeliness constraint change, etc. During model training, the agent observes the current state, calculates the expected return of each action according to the Q-network, and selects the action with the maximum return to execute; after execution, it obtains the reward signal from the environment and observes the new state, and updates the Q-network parameters to make the estimated value approach the real return. The design of the reward function guides the strategy to evolve in the direction of improving the comprehensive security efficiency of the system, and the reward signals include response delay improvement reward, alarm accuracy improvement reward, system resource utilization optimization reward, etc.
[0079] The structure of the Deep Q-Network includes an input layer, several hidden layers and an output layer. The input layer receives the normalized state feature vector, the hidden layer adopts a fully connected network structure, each hidden layer is followed by a ReLU activation function to increase the nonlinear expression ability, the number of nodes in the output layer corresponds to the dimension of the action space, and the output value is the Q-value estimate of the corresponding action. Model training adopts an experience replay mechanism and target network technology. Experience replay stores historical interaction samples in the experience pool, and random sampling disrupts the correlation of samples during training; the target network periodically copies parameters from the main network to maintain the stability of the target value. During the training process, the model parameters are updated through the gradient descent algorithm to minimize the mean square error loss between the predicted Q-value and the target Q-value.
[0080] The strategy parameter update period is set to 24 hours, that is, the system performs a parameter update on the deep Q-network model every 24 hours. During the update, empirical samples accumulated in the past 24 hours are used for batch learning, and the old parameters are replaced with new parameters after the model converges. At the same time, the system records the changes in system performance indicators before and after the strategy parameter adjustment, and evaluates the actual effect of strategy optimization. The strategy optimization process continues, and the model continuously learns from operation data to gradually improve its adaptability to complex environments and sudden security incidents.
[0081] The overall working process of the method of the present invention will be described in detail below by taking a specific application scenario as an example.
[0082] Specific application example: multi-device linkage management and control of the smart park security system A smart park is deployed with an Internet of Things-based security system, which includes five functional modules: a video surveillance subsystem, an access control subsystem, an alarm detection subsystem, a fire protection linkage subsystem, and a perimeter protection subsystem. The video surveillance subsystem includes 50 network cameras, 4 video storage servers and 1 video analysis server. The cameras are connected via the ONVIF protocol, and face recognition and behavior analysis algorithms are deployed on the video analysis server. The access control subsystem includes 30 access controllers, 60 card readers and 2 access control management servers, and the controllers support dual protocol stacks of the Modbus protocol and the OPC UA protocol. The alarm detection subsystem includes 100 infrared detectors, 20 vibration sensors and 1 alarm host, and the alarm host supports the BACnet protocol. The fire protection linkage subsystem includes 50 smoke detectors, 30 temperature-sensitive detectors, 20 spray controllers and 1 fire protection control cabinet, and the controllers adopt the Modbus RTU protocol. The perimeter protection subsystem includes 10 pairs of infrared beam detectors and 1 perimeter alarm host, and the host supports reporting via the MQTT protocol.
[0083] The execution flow of the method of this invention in this scenario is as follows. First, step 1 is executed to build a unified protocol adaptation layer and manage the access of heterogeneous devices. A unified protocol adaptation layer service is deployed, configuring drivers for five protocols: ONVIF, Modbus, OPC UA, BACnet, and MQTT. After the protocol adaptation layer starts, it establishes connections with the main control devices of each subsystem sequentially, negotiating protocol capabilities and adapting data formats. Cameras in the video surveillance subsystem access the adaptation layer via the ONVIF protocol. The adaptation layer receives the RTSP video streams pushed by the cameras and performs protocol conversion, converting them into a unified video stream push format for consumption by upper-layer applications. The access control controller in the access control subsystem supports both Modbus and OPC UA protocols. The protocol adaptation layer selects the corresponding protocol driver for data interaction based on the actual communication method of the controller. The alarm detection subsystem, fire linkage subsystem, and perimeter protection subsystem access the adaptation layer via BACnet, Modbus, and MQTT protocols, respectively. Each device completes the registration process through the device registration module. During device registration, a digital certificate is submitted for identity authentication. After successful authentication, the device capability information is reported and a device identifier assigned by the platform is obtained.
[0084] Then, step 2 is executed to establish a multi-device linkage management platform and device topology relationship management. The centralized management platform activates the device discovery function, scanning the campus network through a combination of active detection and passive monitoring to identify all registered security devices. The device discovery module collects information such as the network address, device type, and manufacturer model of each device to build an initial device list. Based on the device list and network configuration information, the topology building engine automatically constructs a device topology relationship graph, identifying four network areas: video surveillance aggregation switch, access control network switch, fire protection dedicated network, and perimeter protection network, as well as the campus core switch connecting each area. Key devices in the topology graph, including video analytics servers, alarm hosts, and fire control cabinets, are identified as high-importance nodes and configured with enhanced monitoring strategies. Redundant path analysis identifies dual-link redundant connections from the campus core switch to the switches in each area, providing a physical basis for high availability of linkage links.
[0085] Next, step 3 is executed to design a collaborative response timing management mechanism and strategy conflict resolution. Multi-device linkage strategies are configured, including two core strategies: intrusion alarm linkage strategy and fire alarm linkage strategy. The intrusion alarm linkage strategy is configured as follows: when the infrared beam detector of the perimeter protection subsystem detects intrusion, it triggers an alarm on the alarm host, simultaneously linking the video surveillance system to switch the corresponding area's camera feed to the monitoring screen, linking the access control system to lock the relevant area's access control, and linking the perimeter protection subsystem to activate the audible and visual alarms. The priority parameter of the strategy configuration is set to 80, and the time constraint is set to be effective within 24 hours. The fire alarm linkage strategy is configured as follows: when the smoke detector of the fire alarm linkage subsystem detects that the smoke concentration exceeds the threshold, it triggers a fire alarm, simultaneously linking the video surveillance system to switch to the fire area's feed, linking the access control system to open the access control in the fire area and evacuation routes, and linking the alarm detection subsystem to activate the full-area audible and visual alarms. The strategy priority is set to 95, higher than the intrusion alarm linkage strategy. The strategy conflict detection algorithm performs static analysis on the strategy library, confirming that the two strategies do not have overlapping conditions or contradictory actions, and the configuration verification is successful.
[0086] The timing management module generates the execution sequence for the intrusion alarm linkage strategy: Step 1, send an alarm confirmation command to the perimeter alarm host; Step 2, send a screen switching command to the video surveillance management server; Step 3, send an area lock command to the access control controller; Step 4, send a start command to the audible and visual alarm. Each step has a minimum execution interval of 50 milliseconds to ensure timing accuracy.
[0087] Step 4 then executes to implement device status monitoring and fault diagnosis. The status monitoring module collects status data from all registered security devices every 2 seconds. The collected data includes device online status, CPU utilization, memory usage, network connections, video stream bitrate, access control records, and alarm trigger counts. For the video analytics server, additional dedicated metrics such as GPU utilization, video analytics frame rate, and face recognition accuracy are collected. The collected data is stored in a time-series database and sent to the health assessment engine for real-time analysis. The health assessment model comprehensively calculates the health score for each device; under normal operating conditions, the health score for each device remains above 85 points.
[0088] An anomaly detection model continuously monitors the video analytics server's operational data. At a certain point, the model detects a significant drop in the video analytics frame rate, with the reconstruction error exceeding the anomaly threshold, thus determining that the video analytics server has malfunctioned. The fault diagnosis process is initiated. Bayesian network inference identifies a video encoding / decoding module failure as the most likely cause, while a random forest classifier confirms the failure type as software resource exhaustion. Impact assessment analysis determines that the failure affects the video analytics function, potentially causing face recognition and intrusion detection functions to malfunction, and generates a fault propagation path diagram. The system automatically generates a fault report, recommending restarting the video analytics service module or expanding GPU resources. Upon receiving the alarm, maintenance personnel execute the recommended fault handling procedures, and the fault is resolved within 30 minutes.
[0089] Then, step 5 is executed to construct a dynamic scheduling and fault-tolerance mechanism. The dynamic scheduling module continuously monitors the load status of each security device. At a certain moment, the access control management server experiences a surge in load due to batch queries of access control records, with CPU utilization reaching 85%, memory utilization reaching 78%, and task queue length reaching 20. After comprehensive evaluation based on the load weight calculation formula, the server's weight decreases to 0.45. The scheduling module automatically assigns newly generated access control record query tasks to another access control management server for execution, achieving load balancing.
[0090] Key equipment is configured with a primary / backup failover mechanism, and the video analytics server is configured with a three-machine hot standby architecture (1 primary, 2 backups). The primary and backup servers synchronize their states in real time via a dedicated synchronization link, with a heartbeat detection cycle of 500 milliseconds. If the primary video analytics server experiences a hardware failure at a certain time, and the heartbeat detection fails three times consecutively, the system determines that the primary server is unavailable and completes failover within 85 milliseconds. The backup server takes over the business and restores the task context from the synchronized state, ensuring uninterrupted business operations. The task transfer mechanism automatically transfers backlogged video analytics tasks from the primary server to the backup server for continued execution, maintaining a 100% task completion rate.
[0091] The cascading failure prevention algorithm continuously monitors device status. At a certain moment, abnormal packet loss occurred on the video surveillance aggregation switch port, with the packet loss rate rapidly increasing from 0.1% to 15%. After detecting the anomaly, the algorithm analyzed the fault propagation path and predicted that if the switch failed, 30 cameras would go offline, and the intrusion alarm linkage function would fail. The algorithm automatically issued a port shutdown command to isolate the faulty port, triggering a backup link switchover, and camera traffic was switched to the redundant link for transmission. The total time from anomaly detection to link recovery was controlled within 2 seconds, effectively preventing cascading failures.
[0092] Finally, step 6 is executed to construct a security situation assessment and decision optimization system. The situation assessment module performs a comprehensive situation assessment of the park's security system every minute. Assessment data sources include: equipment health data reported by the status monitoring module, alarm event data compiled by the alarm management module, temperature, humidity, and smoke concentration data collected by environmental monitoring sensors, traffic indicator data compiled by the network traffic analysis module, and historical baseline data for each indicator in the historical database. A multi-factor fusion algorithm calculates the comprehensive situation score; the current score is 82 points, and the situation level is determined to be of concern.
[0093] During this period, environmental monitoring data showed that the temperature in the park's perimeter area was slightly higher than the historical average baseline. The deep Q-network model in the linkage strategy optimization module analyzed this trend and determined that the perimeter protection subsystem might require enhanced monitoring. The model's suggested action was to moderately increase the sensitivity parameters of the perimeter infrared beam detectors. After the strategy parameters were updated, the deep Q-network recorded the changes in alarm accuracy and system resource utilization before and after the adjustment as reward signals. After 24 hours of continuous learning and optimization, the model gradually adjusted the perimeter protection strategy parameters to the optimal configuration, increasing the alarm accuracy from 89% to 94% and decreasing the false alarm rate from 11% to 6%.
[0094] The complete implementation process of the method of this invention in a smart park security system is as described above. Seamless access of heterogeneous devices is achieved through a unified protocol adaptation layer; a clear device association view is established through device topology management; the correct execution of coordinated actions of multiple devices is ensured through timing management and conflict resolution; timely detection and accurate location of device anomalies are achieved through status monitoring and fault diagnosis; the continuity and reliability of linked services are guaranteed through dynamic scheduling and fault tolerance mechanisms; and dynamic perception of security situation and adaptive adjustment of policies are achieved through situation assessment and policy optimization.
[0095] Example 2 To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to another specific embodiment.
[0096] This embodiment takes a safety production monitoring system in an industrial manufacturing plant as an example to illustrate in detail the application of the method of the present invention in the industrial Internet of Things environment.
[0097] An industrial manufacturing plant has implemented a safety production monitoring system, encompassing five functional modules: video surveillance, gas monitoring, electrical monitoring, fire alarm linkage, and personnel positioning. The video surveillance subsystem includes 80 industrial-grade network cameras and 2 video storage and analysis servers; the cameras support both ONVIF and GB / T 28181 protocols. The gas monitoring subsystem includes 40 gas detectors and 1 gas monitoring host; the detectors support the Modbus RTU protocol, and the host supports the OPC UA protocol for data reporting. The electrical monitoring subsystem includes 30 electrical fire detectors and 2 electrical monitoring management units; the detectors support the Modbus protocol. The fire alarm linkage subsystem includes 60 smoke and heat detectors and 1 fire control cabinet; the controller uses the BACnet protocol. The personnel positioning subsystem includes 200 positioning tags, 20 positioning base stations, and 1 positioning server; the positioning base stations support the CoAP protocol.
[0098] The execution flow of the method of this invention in this scenario is as follows. First, step 1 is executed to build a unified protocol adaptation layer and heterogeneous device access management. A unified protocol adaptation layer service is deployed, and drivers for six protocols—ONVIF, GB / T 28181, Modbus RTU, OPC UA, BACnet, and CoAP—are configured. Considering the characteristics of a large number of devices and a high degree of protocol heterogeneity in industrial environments, the protocol adaptation layer adopts a distributed deployment architecture. Protocol access front-end machines are deployed in each production workshop. Each front-end machine is responsible for device access and data aggregation in its workshop, and forwards standardized data to the upper-level platform through a message queue.
[0099] The device registration module processes registration requests from various subsystems. When cameras in the video surveillance subsystem register, they submit device certificates for two-way authentication. Upon successful authentication, they report ONVIF capability descriptions, including supported video encoding formats, resolution range, PTZ control capabilities, and other information. The registration process for the gas monitoring host is similar, reporting a list of monitored gas types, detection accuracy parameters, alarm threshold configurations, and other capability information. The device identifier assigner assigns a globally unique device identifier to each device. The identifier encoding rule includes four parts: device type code, manufacturer code, serial number, and check digit, ensuring the uniqueness and verifiability of the identifier.
[0100] Then, step 2 is executed to establish a multi-device linkage management platform and device topology relationship management. Device discovery mechanisms face challenges in industrial network environments due to the large number of devices and complex network structures. The topology building engine adopts a layered construction strategy. First, a workshop-level subnet topology is built to identify the core switches and aggregation switches in each workshop; second, a device-level access topology is built to identify the connection relationships between each device and switch ports; finally, a cross-workshop backbone topology is built to identify the backbone network devices connecting each workshop network. The topology graph adopts a directed weighted graph structure, with edge weights comprehensively considering link bandwidth, transmission latency, and reliability assessment values.
[0101] Constructing a topology map in an industrial environment requires considering the correlation between equipment functional zoning and physical layout. During topology construction, cluster analysis is performed based on the functional attributes and physical locations of the equipment, grouping functionally related and physically adjacent equipment into the same equipment group. Equipment groups serve as the target units for linkage strategy configuration, facilitating regionalized linkage control. For example, gas detectors, video cameras, and audible and visual alarms in a production workshop can be grouped into the same equipment group; when a gas leak alarm is triggered, the equipment within this group will respond in a coordinated manner.
[0102] Next, step 3 is executed to design a collaborative response timing management mechanism and strategy conflict resolution. Multi-device linkage strategies suitable for industrial environments are configured, including three core strategies: gas leak emergency linkage strategy, electrical fire early warning linkage strategy, and personnel intrusion alarm linkage strategy. The gas leak emergency linkage strategy is configured as follows: when the detector of the gas monitoring subsystem detects that the concentration of combustible or toxic gas exceeds the alarm threshold, a gas leak alarm is triggered. Simultaneously, the video monitoring system is linked to retrieve video footage from the leak area and downwind area; the fire protection linkage subsystem is linked to activate the sprinkler system in the accident area; the personnel positioning subsystem is linked to count the number of personnel working in the leak area and issue evacuation instructions; and the audible and visual alarms are linked to issue personnel evacuation alarms. The priority of the strategy configuration is set to 90, and the time constraint is 24 / 7 effectiveness.
[0103] The electrical fire early warning and linkage strategy is configured as follows: when the detectors of the electrical monitoring subsystem detect an abnormal increase in line temperature or an abnormal leakage current, an electrical fire early warning is triggered. Simultaneously, the video surveillance system retrieves video footage from the relevant electrical cabinets, the fire alarm linkage subsystem focuses on monitoring the warning area, and the personnel positioning subsystem confirms the location of nearby maintenance personnel. The strategy priority is set to 85, lower than the gas leak emergency linkage strategy.
[0104] The strategy conflict detection algorithm detected a potential execution conflict risk between the electrical fire early warning linkage strategy and the gas leak emergency linkage strategy: both strategies involve the control of the fire-fighting linkage subsystem, which may result in contradictory actions such as simultaneously controlling the start and stop of the sprinkler system. The conflict resolution module, based on a priority arbitration mechanism, determined that the gas leak emergency linkage strategy should be executed first, and the control actions of the electrical fire early warning linkage strategy on the fire-fighting linkage subsystem were temporarily suspended.
[0105] The timing management module generates a precise execution sequence for the gas leak emergency response strategy, including: Step 1, sending alarm confirmation and concentration reading acquisition instructions to the gas monitoring host; Step 2, sending instructions to the video monitoring management server to retrieve images of the accident area and downwind area; Step 3, sending a sprinkler system activation instruction to the fire control cabinet; Step 4, sending area personnel statistics and evacuation instructions to the personnel positioning server; Step 5, sending an evacuation alarm activation instruction to the audible and visual alarms. A minimum execution interval of 50 milliseconds is set between each step to ensure timing accuracy. The timing management module also maintains a connection with the plant's time synchronization server to ensure that the time deviation of the entire system is controlled within 1 millisecond.
[0106] Step 4 then executes to achieve equipment status monitoring and fault diagnosis. The status monitoring module performs high-density status data collection on equipment in the industrial environment. Considering the special characteristics of the industrial production environment, the data collection cycle is maintained at 2 seconds. Simultaneously, a proactive abnormal status reporting mechanism is configured for critical equipment, reporting immediately when the equipment status changes drastically without waiting for polling. The collected industrial-specific indicators include the calibration status and sensor lifespan warning of gas detectors, the line impedance and insulation resistance of electrical detectors, the lens contamination level and infrared supplementary light status of video cameras, and the positioning accuracy and signal strength of personnel positioning systems, etc.
[0107] The anomaly detection model continuously analyzes the operational data of gas detectors. At a certain moment, the readings of multiple gas detectors exhibited an abnormal pattern of synchronous fluctuations. Although the individual readings did not exceed the alarm threshold, the fluctuation pattern differed significantly from the normal ventilation change pattern. The anomaly detection model determined that there were signs of a slow leak and triggered the diagnostic process. The fault diagnosis process used Bayesian network inference to analyze the fault correlation of multiple detectors, and a random forest classifier determined that the cause of the anomaly was a gas retention response delay caused by a blockage in the gas sampling pipeline. The impact assessment analysis determined that the fault did not affect the alarm function but would affect the accuracy of locating the leak source, and recommended maintenance and repair of the relevant sampling pipeline. Maintenance personnel checked the sampling pipeline according to the recommendation and found a loose pipeline joint. After repair, the detector readings returned to normal.
[0108] Then, step 5 is executed to construct a dynamic scheduling and fault-tolerance mechanism. Industrial environments place extremely high demands on the reliability and continuity of the coordinated control system, necessitating a more stringent fault-tolerance configuration. Key equipment is configured with a three-machine hot standby architecture (1 primary, 2 backups), with millisecond-level switching for primary / backup switching. The heartbeat detection cycle is optimized to 500 milliseconds, and fault switching latency is controlled within 100 milliseconds. The video storage and analysis server, gas monitoring host, and fire control cabinet are all configured in hot standby mode, with backup equipment synchronizing real-time status data with the primary equipment.
[0109] The dynamic scheduling module continuously monitors the load status and health of each device. When it detects that the load of a video storage and analysis server is continuously increasing and approaching a threshold, the scheduling module automatically migrates some video analysis tasks to another server with a lower load, achieving dynamic load balancing. During the task migration process, a state synchronization mechanism ensures the integrity of the task context. The current analysis progress is recorded before migration, and analysis is resumed from the breakpoint after migration, ensuring uninterrupted video analysis services.
[0110] Cascaded failure prevention algorithms prioritize monitoring the status of critical control links in industrial networks. Fire alarm linkage links, as the most crucial safety assurance links, are configured as the highest-priority monitoring target. The algorithm performs in-depth monitoring of these links, real-time detecting end-to-end connectivity from the management platform to the fire control cabinet, the health status of each node, and response latency. If, at a certain moment, the algorithm detects an anomaly of increased packet error rate on an intermediate switch in the fire alarm linkage link, it determines that the switch has a potential failure risk. Before the failure occurs, the algorithm proactively triggers link switching, redirecting fire alarm linkage traffic to redundant paths, thus preventing linkage failure caused by the potential fault.
[0111] Finally, step 6 is executed to construct a safety situation assessment and decision optimization system. Safety situation assessment in industrial environments needs to consider more industry-specific factors, including production operation periods, process stages, hazardous substance storage and distribution, and emergency response plan activation status. The situation assessment module adds industry-specific assessment factors to the standard five-dimensional assessment factors, including process parameter deviation, hazardous substance exposure index, and worker density. The situation classification thresholds are customized according to industrial safety standards, with the boundary score between normal and concern levels adjusted to 85 points, higher than the general scenario setting.
[0112] Assessment data for a certain period showed an increase in the number of alarms from the electrical monitoring subsystem compared to the historical baseline, causing the overall score of the situation assessment module to drop to 78 points, classifying the situation as "concernful." Deep Q-network model analysis of the alarm event sequence revealed that alarms were mainly concentrated in the power distribution room area on the east side of the plant, indicating potential hazards such as aging electrical wiring or overload in this area. The model's action recommendation was to increase the frequency of inspections and the density of electrical parameter monitoring in this area. Maintenance personnel conducted a comprehensive inspection of the area based on the recommendation, discovering a potential faulty connection at a wiring joint, which was promptly repaired, preventing a potential electrical fire.
[0113] The complete implementation process of the method of this invention in the safety production monitoring system of an industrial manufacturing plant is as described above. This application scenario demonstrates the applicability and effectiveness of the method of this invention in the industrial Internet of Things (IIoT) environment. It achieves seamless access for industrial equipment from multiple vendors and using multiple protocols through heterogeneous protocol adaptation; establishes a device association view matching the industrial production process through device topology management; ensures the correctness of multi-device collaborative actions in the industrial environment through time-series management and conflict resolution; achieves early warning and precise location of industrial equipment anomalies through status monitoring and fault diagnosis; ensures the continuity of industrial safety production monitoring services through dynamic scheduling and fault tolerance mechanisms; and achieves dynamic perception of the safety situation and automatic adjustment of preventive maintenance strategies through situation assessment and strategy optimization.
[0114] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.
Claims
1. A multi-device linkage control method for a security prevention system based on the Internet of Things, characterized in that, Includes the following steps: Step 1, Build a unified protocol adaptation layer and heterogeneous device access management: Design and deploy a unified protocol adaptation layer to achieve standardized conversion of communication protocols and data formats of security devices from different manufacturers. Through the device registration module, complete the information collection, capability identification and unified identification allocation of heterogeneous devices, and seamlessly connect heterogeneous security devices to the unified management and control platform. Step 2, establish a multi-device linkage control platform and device topology relationship management: build a centralized control platform to realize unified registration, configuration and management of security devices, automatically build a device topology relationship map through the device discovery mechanism, determine the physical connection and logical association between devices based on the topology map, and provide basic support for linkage strategy configuration; Step 3, Design a collaborative response timing management mechanism and strategy conflict resolution: Establish a timing management module, use priority scheduling algorithm and time synchronization mechanism to ensure the timing correctness of multi-device linkage response, use strategy conflict detection algorithm to identify logical conflicts between linkage rules, use priority arbitration mechanism to resolve conflicts and generate conflict-free linkage execution sequence; Step 4, Implement equipment status monitoring and fault diagnosis: Deploy a status monitoring module to collect the operating status, working parameters and performance indicators of each security device in real time, use a multi-dimensional health assessment algorithm to comprehensively analyze the equipment status, identify equipment faults through anomaly detection models and trigger the diagnosis process, and generate fault reports and maintenance suggestions; Step 5, Construct a dynamic scheduling and fault tolerance mechanism: Establish a dynamic scheduling module to dynamically allocate linkage tasks based on the real-time load and health status of the equipment, adopt a weighted round-robin scheduling algorithm to achieve task load balancing, introduce a primary / backup switching mechanism and a task transfer mechanism to ensure the continuity of linkage services, and prevent fault propagation through a cascading failure blocking algorithm. Step 6: Construct a security situation assessment and decision optimization system: Establish a situation assessment module to comprehensively analyze multi-dimensional data such as equipment status, alarm frequency, and environmental parameters, dynamically assess and classify the overall security situation, and continuously optimize linkage strategy parameters based on historical execution data using reinforcement learning algorithms.
2. The method as described in claim 1, characterized in that: The unified protocol adaptation layer supports more than 20 mainstream IoT communication protocols, is equipped with an extensible protocol registration interface, and supports adding custom protocol drivers through plugins.
3. The method as described in claim 1, characterized in that: The device registration module adopts a two-way digital certificate authentication mechanism to complete protocol capability reporting, functional attribute mapping and resource quota allocation during the device registration process. The protocol conversion module adopts a state machine-driven parsing architecture and realizes asynchronous communication and data buffering through message queues.
4. The method as described in claim 1, characterized in that: The device topology graph is organized using a directed weighted graph structure, where nodes represent security devices, edges represent communication links between devices, and the weights of the edges represent communication latency and reliability assessment values.
5. The method as described in claim 1, characterized in that: The policy conflict detection algorithm performs static analysis and runtime detection on the policy library. Static analysis is performed during policy configuration to detect overlapping conditions and contradictory actions, while runtime detection verifies policy compatibility in the current state before policy execution.
6. The method as described in claim 1, characterized in that: The multi-dimensional health assessment algorithm is designed using the analytic hierarchy process. The model structure includes an indicator layer, a criterion layer, and a target layer. The indicator layer corresponds to various original monitoring indicators, and the criterion layer categorizes the indicators into three evaluation criteria: performance, reliability, and safety.
7. The method as described in claim 1, characterized in that: The anomaly detection model adopts an unsupervised learning algorithm based on autoencoders and achieves anomaly state recognition through training on normal behavior patterns. The fault diagnosis process includes three stages: fault location, fault classification, and impact assessment. Fault location uses Bayesian network inference, and fault classification uses a classifier based on random forest.
8. The method as described in claim 1, characterized in that: The load weight calculation comprehensively considers four factors: CPU utilization, memory utilization, network bandwidth utilization, and task queue length, and uses a weighted summation method to calculate the comprehensive load weight.
9. The method as described in claim 1, characterized in that: The fault tolerance mechanism adopts a primary-secondary redundancy configuration strategy. Key equipment is deployed with primary and secondary dual-machine hot standby. When the primary equipment fails, it automatically switches to the backup equipment. The heartbeat detection cycle is 500 milliseconds, and the fault switching latency is less than 100 milliseconds.
10. The method as described in claim 1, characterized in that: The security situation assessment adopts a multi-factor fusion algorithm, which comprehensively considers five dimensions of assessment factors: equipment operating status, alarm event statistics, environmental monitoring data, network traffic characteristics, and historical statistical patterns. The situation classification divides the security situation into five levels: normal, attention, alert, threat, and severe.
