Internet of Things equipment identification method and system based on lightweight knowledge graph

Abstract

The invention belongs to the field of Internet of Things equipment identification, and provides an Internet of Things equipment identification method and system based on a lightweight knowledge graph, and the method comprises the steps: carrying out the preliminary scanning to obtain the data information of multiple dimensions of Internet of Things equipment and the equipment type, and obtaining a comprehensive feature; constructing a knowledge graph by taking the equipment type as a central node, taking each piece of data of each dimension as a node and taking a corresponding relationship between each node and the central node as an edge; continuously identifying the Internet of Things equipment to determine the corresponding latest comprehensive characteristics, and calculating the similarity between the latest comprehensive characteristics and the comprehensive characteristics in the knowledge graph to generate a candidate equipment type list; verifying the device types in the candidate device type list by using the context information, if the verification is successful, determining the final matched device type, and updating the latest comprehensive characteristics to the device type corresponding to the knowledge graph; otherwise, updating the knowledge graph according to the newest comprehensive features as new nodes.

Description

Technical Field

[0001] The present invention belongs to the technical field of Internet of Things device identification, and specifically relates to an Internet of Things device identification method and system based on a lightweight knowledge graph. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid development of Internet of Things (IoT) technology, the number of IoT devices has exploded, covering a variety of types such as sensors, cameras, and smart home devices, and their functions and application scenarios have become increasingly complex and diverse. Accurately and quickly identifying these devices is not only an important means to improve network security management, but also the key to achieving optimal resource allocation. Traditional device identification methods usually rely on static information acquisition, such as IP addresses and MAC addresses, which are difficult to adapt to dynamically changing network environments. This method is particularly weak in the face of constantly changing device status and configuration, and is prone to identification delays or errors, which in turn affects the security and management efficiency of the entire network. In order to meet this challenge, the development of modern technology has made it possible to combine advanced network scanning tools and intelligent knowledge graph construction.

[0004] In the prior art, although the method of using knowledge graphs for IoT device identification has advantages in representing complex relationships and rich semantics, it also has some defects and shortcomings. First, the construction and maintenance costs of knowledge graphs are high, requiring a lot of data cleaning and professional knowledge. Especially when facing a variety of IoT devices, data updates and accuracy are difficult to guarantee. Secondly, knowledge graphs rely on historical data and prior knowledge, and it is difficult to respond to dynamic changes in device status in a timely manner, resulting in poor real-time performance. In addition, in a large-scale network environment, the reasoning efficiency of the knowledge graph may become a bottleneck, making it difficult to meet the needs of efficient identification and response. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a method and system for identifying IoT devices based on lightweight knowledge graphs. The present invention combines the fusion of IoT device-related features, knowledge graph construction, similarity matching and context verification to efficiently and accurately identify and classify various IoT devices. At the same time, the dynamic update and optimization mechanism ensures the adaptability and long-term stability of the system, enabling it to cope with future changes in device types and features.

[0006] According to some embodiments, a first solution of the present invention provides an IoT device identification method based on a lightweight knowledge graph, which adopts the following technical solutions:

[0007] A method for identifying IoT devices based on a lightweight knowledge graph, comprising:

[0008] Obtain data information and device types of multiple dimensions of IoT devices, calculate and sort the mutual information values ​​between the data information and device types of multiple dimensions, select the data information of the corresponding dimensions with mutual information values ​​greater than a set threshold and fuse them with the device type to obtain comprehensive features;

[0009] Take the device type as the central node, each data in each dimension as a node, and the corresponding relationship between each node and the central node as an edge to build a knowledge graph;

[0010] Continue to identify data information and device types of multiple dimensions of IoT devices, determine the corresponding latest comprehensive features, and calculate the similarity between the latest comprehensive features and the comprehensive features in the knowledge graph, and generate a list of candidate device types based on the device types corresponding to the comprehensive features whose similarity is greater than the similarity threshold;

[0011] The context information is used to verify the device type in the candidate device type list. If the verification is successful, the final matching device type is determined, and the latest comprehensive features are updated to the corresponding device type in the knowledge graph; if the verification fails, the knowledge graph is updated based on the latest comprehensive features as a new node.

[0012] Furthermore, the data information of multiple dimensions includes device ID, device name, device manufacturer, device IP address, device communication protocol, device port status, device location, device working scenario, device installation time, device maintenance record and device behavior.

[0013] Furthermore, the use of context information to verify the device type in the candidate device type list is specifically as follows:

[0014] According to the candidate device type list, obtain the context node related to the device type;

[0015] The context node of the IoT device to be verified is verified for consistency with each type in the candidate device type list. If the verification is successful, the final matching device type is determined; otherwise, the knowledge graph is updated based on the latest comprehensive features as a new node.

[0016] Furthermore, the context nodes include a device location node, a device time node, a device maintenance record node, and a device behavior node.

[0017] Furthermore, the consistency verification of the context node of the IoT device to be verified with each type in the candidate device type list is specifically as follows:

[0018] Condition 1: Whether the device location node of the IoT device to be verified matches each type in the candidate device type list;

[0019] Condition 2: Whether the typical behavior in the device behavior node of the IoT device to be verified matches each type in the candidate device type list;

[0020] Condition 3: If the device installation time node or device maintenance record node of the IoT device to be verified is obtained, then determine whether it matches each type in the candidate device type list; otherwise, the device installation time node and device maintenance record node are not verified;

[0021] If at least one condition is met, the consistency verification succeeds.

[0022] Furthermore, if there are multiple successfully verified candidate device types, the device type with the highest similarity is selected as the final matching device type.

[0023] Furthermore, if the similarity between the latest comprehensive feature and the comprehensive feature in the knowledge graph is less than a similarity threshold, the knowledge graph is updated based on the latest comprehensive feature as a new node.

[0024] According to some embodiments, a second solution of the present invention provides an IoT device identification system based on a lightweight knowledge graph, which adopts the following technical solutions:

[0025] An IoT device identification system based on a lightweight knowledge graph, comprising:

[0026] The preliminary identification module is configured to obtain data information of multiple dimensions and device types of IoT devices, calculate and sort the mutual information values ​​between the data information of multiple dimensions and the device types, select the data information of the corresponding dimensions with a mutual information value greater than a set threshold and fuse them with the device type to obtain a comprehensive feature;

[0027] The knowledge graph construction module is configured to construct the knowledge graph by taking the device type as the central node, taking each data of each dimension as a node, and taking the corresponding relationship between each node and the central node as an edge;

[0028] The knowledge graph matching module is configured to continue to identify data information of multiple dimensions and device types of IoT devices, determine the corresponding latest comprehensive features, calculate the similarity between the latest comprehensive features and the comprehensive features in the knowledge graph, and generate a candidate device type list based on the device types corresponding to the comprehensive features whose similarity is greater than the similarity threshold;

[0029] The context verification module is configured to use context information to verify the device type in the candidate device type list. If the verification is successful, the final matching device type is determined and the latest comprehensive features are updated to the corresponding device type in the knowledge graph; if the verification fails, the knowledge graph is updated based on the latest comprehensive features as a new node.

[0030] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium.

[0031] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for identifying an Internet of Things device based on a lightweight knowledge graph as described in the first aspect above.

[0032] According to some embodiments, a fourth aspect of the present invention provides a computer device.

[0033] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for identifying an Internet of Things device based on a lightweight knowledge graph as described in the first aspect above are implemented.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] After acquiring the multi-dimensional information of IoT devices, the present invention adopts mutual information optimization feature selection technology to cope with the multi-dimensional information of a large number of devices. By calculating the mutual information between the multi-dimensional information of IoT devices and the device type, the most discriminating features are screened out, thereby improving recognition efficiency and accuracy. Combined with the similarity calculation strategy, in the process of IoT device identification, when a new unknown device appears, its feature information is matched with the comprehensive feature vector of the known device in the knowledge graph to calculate the similarity, and the device type is matched according to the similarity score and the preset threshold, and a candidate type list is generated. The recognition accuracy is further improved through context verification, and the actual type of the device is confirmed by combining the dynamic information such as the deployment environment of the device and historical maintenance records. The system can efficiently and accurately identify and classify various IoT devices.

[0036] The present invention effectively reduces the computational and storage overhead of the knowledge graph by constructing a lightweight IoT knowledge graph, making it more suitable for IoT environments with limited resources. By adopting the dynamic update of real-time data streams and the on-demand loading technology of data slicing, the system can update the characteristics and behaviors of the device at any time, and only load the device data related to the current task, thereby significantly improving the storage efficiency and response speed. At the same time, combined with the context verification mechanism (including the deployment environment, behavior patterns, historical information, etc. of the device), the system can further reduce device recognition errors and make the recognition results more accurate and in line with actual application scenarios. Ultimately, these optimization measures greatly enhance the real-time, flexibility and adaptability of the system in the IoT device recognition process. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0038] Figure 1 is a flow chart of an IoT device identification method based on a lightweight knowledge graph in an embodiment of the present invention;

[0039] Figure 2 Schematic diagram of the IoT device identification process in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0041] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0042] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0043] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0044] Terminology explanation:

[0045] IoT devices: Internet of Things (IoT) devices are smart devices that can communicate with each other and exchange data through the Internet. These devices include smart home devices, smart cameras, smart watches, health trackers, industrial control systems, etc. The popularity of IoT devices has brought great convenience, but it is also accompanied by security threats.

[0046] Knowledge graph: Knowledge graph is a way to represent and store knowledge through a graph structure, which is used to describe entities and their relationships. It consists of nodes (representing entities) and edges (representing relationships), allowing information to be organized and queried in an intuitive and structured way. Nodes represent objects in the real world, such as people, places, things or concepts, and edges represent the relationships between these objects. Knowledge graphs are widely used in fields such as natural language processing, recommendation systems, search engines, and intelligent question and answer to help understand semantics and provide personalized recommendations. Building a knowledge graph usually includes steps such as data collection, cleaning, entity recognition, relationship extraction, and knowledge fusion. With the development of artificial intelligence and big data technologies, knowledge graphs have become an important part of intelligent systems.

[0047] Nmap: Nmap (Network Mapper) is an open source network scanning and security auditing tool, widely used in network security and network management. It is mainly used to discover devices, services and their configuration status in the network. It can detect the status of the target host by sending specific network data packets, including open ports, running services and operating system information. Nmap has a host discovery function, which can identify online devices in the network and supports multiple scanning methods, such as Ping scanning and ARP scanning. At the same time, it can detect the open ports of the target host to determine which services are running, and supports multiple protocols such as TCP, UDP and SCTP. In addition, Nmap can also fingerprint the services of the open ports, determine the type of service and its version information, and analyze the response data to infer the operating system and version running on the target host. Nmap provides a powerful scripting engine (NSE), which allows users to write and run scripts to automate various network tasks, such as vulnerability scanning and security assessment. Its flexibility and scalability make Nmap an indispensable tool for network administrators and security experts, and it is widely used in scenarios such as network security auditing, vulnerability assessment and network monitoring.

[0048] IoT Knowledge Graph (IoT-KG): IoT-KG is mainly aimed at the entire IoT ecosystem, which includes not only sensor data, but also devices, communication protocols, service layer information and the interconnection between them. It can represent a more comprehensive IoT scenario, not only limited to sensors, but also edge devices, cloud computing nodes, etc., and is suitable for scenarios that require unified management of multiple types of devices in the IoT.

[0049] Resource Description Framework (RDF): Resource Description Framework (RDF) is a standardized metadata model proposed by W3C (World Wide Web Consortium), mainly used to represent and exchange information on the Internet. The core idea of ​​RDF is to describe the relationship between resources in the form of triples. Each triple consists of three parts: subject, predicate and object. Among them, the subject represents the resource or object of the data, the predicate describes the relationship between the subject and the object, and the object can be another URI or literal value (such as a number, string, etc.) of the resource. Through this structure, RDF can flexibly represent and connect the relationship between different resources, thereby realizing the sharing and reuse of knowledge.

[0050] Mutual Information: Mutual Information (MI) is a concept used to measure the degree of information correlation between two random variables. Simply put, mutual information indicates how much information one variable contains about another variable. It reveals the dependency between variables rather than a linear relationship. In IoT device identification, mutual information can be used to measure the information dependency between a certain scanning feature (such as open ports, protocol types, etc.) and the device type. For example, if we choose a feature with a high mutual information value, it means that the feature provides more effective information about a specific device type, which helps to identify or distinguish device types.

[0051] Embodiment 1

[0052] like Figure 1 As shown, this embodiment provides an IoT device identification method based on a lightweight knowledge graph. This embodiment uses the method applied to a server as an example. It can be understood that the method can also be applied to a terminal, and can also be applied to a terminal, a server, and a system, and is implemented through the interaction between the terminal and the server. The server can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communications, middleware services, domain name services, security services CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, which is not limited in this application. In this embodiment, the method includes the following steps:

[0053] Step S1: Acquire data information of multiple dimensions and device types of IoT devices, calculate and sort the mutual information values ​​between the data information of multiple dimensions and the device types, select the data information of the corresponding dimensions with mutual information values ​​greater than a set threshold and fuse them with the device types to obtain comprehensive features;

[0054] Step S2: Take the device type as the central node, each data in each dimension as a node, and the corresponding relationship between each node and the central node as an edge to construct a knowledge graph;

[0055] Step S3: Continue to identify data information and device types of multiple dimensions of IoT devices, determine the corresponding latest comprehensive features, and calculate the similarity between the latest comprehensive features and the comprehensive features in the knowledge graph, and generate a list of candidate device types based on the device types corresponding to the comprehensive features whose similarity is greater than the similarity threshold;

[0056] Step S4: Use context information to verify the device type in the candidate device type list. If the verification is successful, determine the final matching device type and update the latest comprehensive features to the corresponding device type in the knowledge graph. If the verification fails, update the knowledge graph based on the latest comprehensive features as new nodes.

[0057] like Figure 2 As shown, step S1: Nmap preliminary scan

[0058] The data information in the multiple dimensions includes device ID, device name, device manufacturer, device IP address, device communication protocol, device port status, device location, device working scenario, device installation time, device maintenance record and device behavior.

[0059] First, use the Nmap tool to scan the target network to obtain the device IP address, device port status, and device communication protocol. This information, as the device's network information, lays the foundation for further identification and classification of subsequent devices. At the same time, a series of information such as device ID, device name, device manufacturer, device location, device working scene, device installation time, device maintenance record, and device behavior are also obtained. The above information and the device's network information are used as data information of multiple dimensions of IoT devices.

[0060] In order to cope with the multi-dimensional information of massive IoT devices, mutual information is introduced to optimize the feature selection and scanning process. By calculating the mutual information between data information in multiple dimensions and device types, the key features shared by each device can be identified, that is, the common information of devices with the greatest identification value. These common features can reveal significant patterns and differences in device types, thereby making the scanning process more focused on effective information. The features are sorted according to the mutual information value, and those with higher mutual information values ​​are selected as the information to be focused on. These features can provide more useful information about the device type and help us identify different IoT devices more accurately. Specifically, the calculation of mutual information can screen out features with high information gain on device categories, such as certain specific port configurations or service protocols, and remove redundant information that has little impact on identification. Finally, the core features with high mutual information are sampled, which not only improve the efficiency and accuracy of IoT device identification, but also realize the lightweight data structure in the device identification model.

[0061] The mutual information values ​​between data information of multiple dimensions and device types are calculated and sorted, and the data information of the corresponding dimensions with mutual information values ​​greater than the set threshold are selected for fusion with the device type. A multi-layer perceptron (MLP) model is used for feature fusion, but not limited to, to generate a comprehensive feature Z as the core feature representation of the sensor node in the knowledge graph. The construction of this feature ensures that the graph contains the functional characteristics and communication information of the device, laying the foundation for subsequent sensor matching.

[0062] Step S2: Knowledge Graph Construction - IoT-KG

[0063] Through the knowledge graph, IoT devices in the network can be discovered and identified by matching device models, protocols and behavioral characteristics.

[0064] First, define the data format of the knowledge graph. Then, design a knowledge graph representation model to convert the structured knowledge graph into a learnable vector representation for downstream learning tasks. For the "IoT device discovery method based on lightweight knowledge graph", IoT-KG (Internet of Things Knowledge Graph) is used to realize the discovery of IoT devices. The following description shows the data format of IoT-KG, the definition of nodes and edges, the embodiment of context information, and how to implement lightweight strategies.

[0065] Data format selection:

[0066] The three existing data formats: Resource Description Framework (RDF), Web Ontology Language (OWL) and Linked Open Data (LOD), each has different characteristics and advantages in knowledge modeling.

[0067] RDF is a general data description framework that describes resources and their relationships through a simple triple structure. It is very suitable for building basic knowledge graphs. OWL provides stronger semantic expression and reasoning capabilities based on RDF, and is suitable for building complex ontology structures and expressing complex logical relationships. LOD emphasizes the links and openness between data, allowing knowledge distributed in different data sources to be interconnected, and is an important technology for building a global shared knowledge network. When building the Internet of Things Knowledge Graph (IoT-KG), RDF is the most suitable choice, especially for the lightweight knowledge graph requirements. First of all, the triple structure of RDF is simple and flexible, which can effectively describe the basic information, attributes and relationships of IoT devices, and meet the basic requirements of device discovery and modeling of relationships between devices in the IoT environment. The lightweight nature of RDF makes it not generate too much computing and storage overhead when processing large-scale IoT device data, which is suitable for the IoT environment with limited resources. In contrast, although OWL is more powerful in describing complex relationships and supporting semantic reasoning, its reasoning mechanism is relatively complex and the computational cost is high. It is not suitable for application in scenarios such as IoT-KG that require lightweight and fast response. Although the openness and data interconnection capabilities of LOD help enhance the interoperability of IoT-KG and its links with other knowledge graphs, its implementation and management are relatively complex, especially when processing real-time IoT data, the data governance and mapping issues it brings also increase the difficulty of implementation. Therefore, in the construction of lightweight knowledge graphs, RDF can provide sufficient descriptive capabilities while maintaining the simplicity and efficiency of the model, and is the best choice for IoT device discovery and identification. At the same time, RDF also has good interoperability, which can lay the foundation for future expansion.

[0068] The main components of IoT-KG knowledge graph are triples (subject-predicate-object), which can be used to express the relationship between entities. The data format of IoT-KG is represented by RDF (Resource Description Framework) and managed by GraphDB. It is understandable that other data formats can also be selected, and the selection of data format is not the core point of protection.

[0069] The following is how IoT-KG data is represented:

[0070] -Nodes: Nodes in IoT-KG represent entities in the network, such as device type, device communication protocol, device manufacturer, etc.

[0071] -Edges: Edges represent the relationship between nodes, such as "connected using a protocol", "manufactured by a manufacturer", "deployed in a certain environment", etc.

[0072] Data format example:

[0073] Data is usually represented as a triple structure similar to the following:

[0074] ```

[0075] <device_1> <type><IoT_Sensor>.

[0076] <device_1> <hasprotocol> <http>.

[0077] <device_1> <location><Room_101>.

[0078] <device_1> <manufacturer><Brand_A>.

[0079] ```

[0080] <device_1> It is a subject, which represents a specific device.

[0081] <type> 、 <hasprotocol> 、 <location>It is a predicate, which represents the attribute or relationship of a device.

[0082] <IoT_Sensor> , <http>,<Room_101> ,<Brand_A> An object represents an attribute value or entity associated with a subject.

[0083] In this embodiment, the device type is used as the central node, each data in each dimension is used as a node, and the corresponding relationship between each node and the central node is used as an edge to construct a knowledge graph.

[0084] Each data of each dimension is regarded as a node, including device ID node, device name node, device manufacturer node, device IP address node, device communication protocol node, device port status node, device location node, device working scenario node, device installation time node, device maintenance record node and device behavior node.

[0085] Device type node: for example, device type (sensor, camera, smart light bulb), device model, manufacturer, device function, etc.

[0086] Device communication protocol node: the communication protocol used by the device (such as HTTP, MQTT, CoAP).

[0087] Device location node: The physical deployment location of the device (such as "Office 101").

[0088] Device behavior node: Typical behavior patterns of devices, such as periodic temperature reporting, timed video recording, etc.

[0089] Edge definition:

[0090] Edges represent the relationship between different nodes, for example:

[0091] -Device-Protocol Relationship: describes the communication protocols supported by the device. For example:<device_1> <supports> <http>.

[0092] -Device-Manufacturer Relationship: describes which manufacturer manufactured the device. For example:<device_1> <manufacturedby><Brand_A>.

[0093] -Device-behavior relationship: describes the behavioral characteristics of a device.

[0094] For example:<device_1> <exhibitsbehavior><periodic_data_report>

[0095] -Device-location relationship: describes the physical location where the device is deployed. For example:<device_1> <locatedin><Room_101>.

[0096] The above are just examples, and the corresponding edges can be determined according to actual needs.

[0097] Step S3: IoT-KG matching:

[0098] Continue to identify data information and device types of multiple dimensions of IoT devices, determine the corresponding latest comprehensive features, and calculate the similarity between the latest comprehensive features and the comprehensive features in the knowledge graph, and generate a list of candidate device types based on the device types corresponding to the comprehensive features whose similarity is greater than the similarity threshold.

[0099] After the knowledge graph is constructed, when a new unknown sensor device appears in the network, the latest feature vector of the IoT device is obtained according to the process of step S1. The latest feature vector The similarity is calculated with the feature vector Z of the known sensors in the knowledge graph to determine the matching degree of the device with various known IoT devices.

[0100] In the process of IoT device identification, accurate device matching and efficient similarity calculation are the key to improving identification accuracy and efficiency. The most suitable calculation method can be selected according to the characteristics of different types of devices to ensure that the device identification system can show high accuracy and efficiency in different scenarios.

[0101] Weighted Cosine Similarity

[0102] For situations where fine-grained comparison is required in device features, such as device port information, protocol type, manufacturer information, etc., weighted cosine similarity is used to calculate the similarity between devices. Traditional cosine similarity calculation is based on the angle between vectors and is usually used for text matching or other high-dimensional sparse data, but it cannot well reflect the importance of different features. To make up for this deficiency, a weighted strategy is introduced to assign different weights to different features according to their importance in device identification. For example, in some devices, manufacturer information may be more important, while in other devices, port information may have a higher degree of discrimination. Therefore, weighted cosine similarity can more accurately reflect the actual correlation between features and improve the accuracy of fine-grained device comparison.

[0103] The weighted cosine similarity formula is as follows:

[0104]

[0105] Among them, fi(A) and fi(B) represent the values ​​of devices A and B on feature i, and Wi is the weight of the feature.

[0106] Similarity calculation based on deep learning

[0107] For devices with complex features and high dimensions (such as devices with multiple sensors, processors, and complex network protocols), traditional similarity calculation methods may face problems such as low computational efficiency or difficulty in capturing the underlying complex relationships between features. To address this problem, we use deep learning models, especially neural networks (such as deep autoencoders or convolutional neural networks), to automatically learn the similarities between device features.

[0108] Deep learning models can automatically learn the nonlinear relationship between device features from a large amount of data and obtain more accurate feature representations through training. These representations can not only be effectively stored in a lower dimension, but also automatically extract more effective feature representations when matching devices, thereby significantly improving the matching accuracy.

[0109] For example, using a deep autoencoder to perform unsupervised learning of device features, by compressing the high-dimensional features of the device into a low-dimensional representation, the deep learning model can retain important semantic information while ignoring irrelevant noise. This low-dimensional embedded feature can more effectively perform similarity comparison.

[0110] Locality Sensitive Hashing (LSH)

[0111] For scenarios that require fast similarity calculation, especially when there are a large number of devices, traditional similarity calculation methods may not be real-time due to the large amount of calculation. Therefore, we introduced the Locality Sensitive Hashing (LSH) technology to speed up the similarity calculation. LSH maps high-dimensional features to low-dimensional space, which can map similar device features to the same hash bucket, reducing the amount of calculation when calculating the distance.

[0112] LSH is particularly suitable for high-dimensional data and can maintain high efficiency on large-scale data sets, avoiding the inefficient practice of brute-force calculation of the similarity of each pair of devices. By properly designing the hash function, LSH can quickly find similar devices and provide real-time support for device identification.

[0113] Combining the above three similarity calculation strategies, the system can flexibly select the most suitable calculation method according to the different types of device features. In practical applications, the features of some devices are relatively simple, which is suitable for fine matching using weighted cosine similarity; for high-dimensional and complex device features, deep learning models can automatically extract more meaningful features and provide more efficient and accurate identification; when the number of devices is large and fast calculation is required, LSH can provide efficient acceleration support.

[0114] In order to improve the accuracy and efficiency of device identification, the strategy fusion method can be further adopted to perform weighted combination of multiple similarity calculation results. By weighted averaging or voting the results of different strategies, the system can improve the accuracy of the final match while ensuring efficiency. For example, if the results of weighted cosine similarity and deep learning models are relatively consistent, the confidence of the matching results can be improved; and when encountering approximate results generated by LSH, further verification can be performed through contextual information or other features to reduce the risk of misidentification.

[0115] The similarity score is calculated based on the latest comprehensive features and the comprehensive features of each candidate device type in the knowledge graph, and the similarity between the device and each IoT device type is quantified using the similarity calculation method mentioned above. For example, the system may derive a similarity score between the IoT device to be matched and a temperature sensor, humidity sensor, or environmental monitoring sensor. The similarity score of each type indicates the degree of match between the device and that type, and the system generates a matching probability for each candidate type, reflecting the strength of the similarity between the device and that type.

[0116] Set a similarity threshold θ. If the similarity score between the IoT device to be matched and a certain type of device does not reach the threshold, the device is considered not to belong to that type. This mechanism helps to avoid false matches and ensures that the IoT device to be matched can only be classified as the type that is closest to its characteristics.

[0117] A candidate device type list is generated based on the device types corresponding to the comprehensive features with similarities greater than the similarity threshold. The candidate device type list lists multiple types that the device is most likely to belong to, and a similarity score is attached to each candidate type. These scores can not only provide preliminary device type recognition results, but also provide a reference for subsequent context verification and recognition accuracy optimization, helping the system to further improve recognition accuracy.

[0118] If the similarity between the latest comprehensive feature and the comprehensive feature in the knowledge graph is less than the similarity threshold, the knowledge graph is updated based on the latest comprehensive feature as a new node.

[0119] Step S4: Context verification.

[0120] In the process of IoT device identification, the context verification mechanism is an important step to improve the accuracy and reliability of identification. Combined with context information, it can further accurately verify the identification results of device types and reduce the probability of misidentification, especially when there are similar features or overlapping working modes between devices.

[0121] The goal of the contextual verification mechanism is to use the dynamic attributes of the device, such as environmental information, usage scenarios, and behavioral characteristics, to further proofread and confirm the static characteristics of the device (such as device type, communication protocol, port configuration, etc.). This process enhances the accuracy of device identification by searching for context nodes related to the device in the knowledge graph (such as location nodes, usage scenario nodes, time nodes, etc.).

[0122] During the matching process, the system not only relies on the static features of the device (such as device type, communication protocol, etc.), but also verifies it in combination with its dynamic context information. This process first confirms the actual situation of the device by analyzing the context nodes corresponding to the device in the knowledge graph. For example, if the identified device type is "temperature sensor", the system will further verify the context information such as the installation location and installation time of the device to ensure that the recognition result meets expectations.

[0123] The use of context information to verify the device type in the candidate device type list is specifically as follows:

[0124] According to the candidate device type list, obtain the context node related to the device type;

[0125] The context node of the IoT device to be verified is verified for consistency with each type in the candidate device type list. If the verification is successful, the final matching device type is determined; otherwise, the knowledge graph is updated based on the latest comprehensive features as a new node.

[0126] The context nodes include device location nodes, device time nodes, device maintenance record nodes, and device behavior nodes.

[0127] Verify the consistency of the context node of the IoT device to be verified with each type in the candidate device type list, specifically:

[0128] Condition 1: Whether the device location node of the IoT device to be verified matches each type in the candidate device type list;

[0129] Condition 2: Whether the typical behavior in the device behavior node of the IoT device to be verified matches each type in the candidate device type list;

[0130] Condition 3: If the device installation time node or device maintenance record node of the IoT device to be verified is obtained, then determine whether it matches each type in the candidate device type list; otherwise, the device installation time node and device maintenance record node are not verified;

[0131] If at least one condition is met, the consistency verification succeeds.

[0132] If there are multiple candidate device types that are successfully verified, the device type with the highest similarity is selected as the final matching device type.

[0133] Specifically, the system will verify based on the following contextual information:

[0134] (1) Equipment deployment environment: The physical location of the equipment is crucial to confirming the equipment type. For example, temperature sensors are usually installed in environments with large temperature changes, such as offices, warehouses, and computer rooms. If a device is identified as a temperature sensor, but its actual deployment location is an outdoor environment or a place with high humidity, the system will determine that the device may not be a temperature sensor, but may be a humidity sensor or environmental sensor. Therefore, the system will use the device's installation location information as a key verification basis.

[0135] (3) Time information: In addition to the static installation location, time factors also play an important role in contextual verification. For example, the device's service life, installation time, or recent maintenance records can all be part of the contextual information. When the system identifies a device as a temperature sensor, if the device's installation time or the last maintenance time does not match its type, the system will correct the device's identification results based on historical information. For example, if a sensor has experienced multiple failures in a short period of time, its working status and properties may need to be re-evaluated.

[0136] (4) Device interaction records and operation logs: In addition to the static properties of the device itself and the deployment environment, user interaction information can also provide important contextual information. By checking the interaction records between the device and the user (such as switch operations, parameter adjustments, error logs, etc.), the system can determine the actual function of the device. For example, after a device is identified as a temperature sensor, if its log shows frequent state changes and user control operations, the system can use this information to confirm whether the device is a sensor that supports temperature control (such as a heater or air conditioner) rather than a single temperature sensor.

[0137] The implementation process of the context verification mechanism usually includes the following steps:

[0138] Contextual information association: After initially identifying the candidate device type, the system will query the knowledge graph to obtain contextual information related to the device. This information includes the physical location of the device (such as an office or computer room), historical maintenance records (such as device installation time and fault repair status), and behavior patterns (such as device communication frequency and interaction behavior). Through this contextual information, the system can have a more comprehensive understanding of the device's working environment and actual usage scenarios.

[0139] Context consistency verification: Compare the device's context information with each type in the candidate device type set to check whether it is consistent with the device's actual environment. For example, if the candidate set contains two types, "temperature sensor" and "humidity sensor", but the device is deployed in a room with high humidity and the device's behavior pattern meets the characteristics of humidity monitoring, the system will exclude the "temperature sensor" type based on the context information and retain the "humidity sensor" as the final device type.

[0140] Verification steps:

[0141] ① Deployment environment consistency: Check whether the installation location of the device is consistent with the candidate type. For example, temperature sensors are usually deployed in areas with large temperature changes, while humidity sensors are deployed in environments with large humidity changes.

[0142] ②Behavior pattern matching: Verify whether the device's behavior pattern matches the typical behavior of a sensor. For example, a temperature sensor generally uploads temperature data at a fixed time, while a humidity sensor may report data at a different frequency.

[0143] ③ Time information verification: If the maintenance history or installation time of the equipment matches a certain sensor type.

[0144] By introducing the contextual verification mechanism, the IoT device identification system can not only rely on traditional static device features, but also perform intelligent verification based on dynamic contexts such as the device's environment, behavior patterns, and historical information. This mechanism greatly enhances the accuracy, reliability, and flexibility of device identification, and provides a solid technical guarantee for the efficient operation of the intelligent IoT system.

[0145] Equipment classification and identification:

[0146] Through the similarity matching and context verification mechanisms mentioned above, the system can accurately classify new devices into appropriate sensor types. For example, the system can identify and classify new devices as temperature and humidity sensors, smoke sensors, motion sensors, etc., and the specific classification is based on the similarity of device features and business needs. The similarity matching process classifies new devices based on a predefined set of device features (such as port configuration, communication protocol, operating frequency, etc.), while the context verification further confirms the accuracy of the classification based on the actual usage scenario, deployment environment, and functional requirements of the device.

[0147] Once a new device is accurately classified, the system will assign a unique identifier to the device and integrate it into the knowledge graph. This identifier not only serves as the unique identification code of the device, but also forms a relationship mapping with other related devices, reflecting the interaction and dependency between devices through connections with other nodes and edges. Specifically, each device label includes but is not limited to the following information:

[0148] -Device type: Specify which type of sensor the device belongs to (such as temperature and humidity sensor, pressure sensor, image sensor, etc.), and provide a basic description of the device's functions and uses.

[0149] -Manufacturer information: Identifies the manufacturer of the device, which helps with device traceability and management, especially in device maintenance and compatibility analysis.

[0150] - Installation location: Based on the location information when the device is deployed (such as building, area, level, specific room, etc.), it helps to further optimize device scheduling, monitoring and troubleshooting.

[0151] - Functional description: Provides detailed functional parameters of the device, including operating mode, data output type, supported protocols, etc., to facilitate functional identification and management during device use.

[0152] - Equipment status and history: Such as the equipment's online / offline status, operating hours, fault history, update records, etc. This information provides basic data for equipment lifecycle management.

[0153] Through these detailed device tags, the system can not only accurately manage and monitor the equipment, but also conduct in-depth analysis and decision support. For example, when a device fails or its performance degrades, the system can intelligently dispatch and arrange relevant maintenance personnel based on the manufacturer information and failure history of the device; in the process of aggregating and analyzing device data, the system can optimize data processing and storage strategies based on the installation location and functional description.

[0154] In addition, the dynamic update of device tags and knowledge graphs provides support for device lifecycle management, data analysis, and predictive maintenance. This process makes the entire IoT system more intelligent, able to continuously adjust strategies based on real-time changes and user needs, and ensure the efficiency and reliability of the system.

[0155] The addition of new devices and the dynamic expansion of the knowledge graph:

[0156] During the device matching process, when the similarity of the unknown sensor does not reach the preset threshold (θ), the system determines that the device is of a new type. At this time, the system creates a new device node in the knowledge graph based on the feature information of the new device (such as model, manufacturer, protocol type, behavior pattern, etc.), and constructs related attribute nodes (such as device function, communication protocol, installation environment, etc.). This process realizes the dynamic expansion of the device type library, ensuring that the system can continuously adapt to the addition of new devices.

[0157] Through the dynamic expansion mechanism, the knowledge graph can be updated in a timely manner to ensure the system's ability to identify new devices. This mechanism supports the discovery and reasoning of device types, keeping the graph up to date, thereby improving the accuracy of device identification and the intelligence of management.

[0158] Continuously optimize feedback mechanism

[0159] During the IoT device identification process, the system continuously optimizes the matching algorithm and knowledge graph update strategy by collecting user feedback and usage data. Users submit feedback on device identification results through the management interface, including recognition accuracy, relevance of device function recommendations, etc. This feedback information will be analyzed through machine learning and natural language processing technology to identify recognition errors and common problems, especially frequent misidentification of a certain type of device. Based on the feedback analysis, the system will adjust the feature weights of the recognition model or retrain the algorithm to improve recognition accuracy.

[0160] At the same time, the system will automatically update the knowledge graph based on the feedback, adjust the device's attributes, behavior patterns, or relationships between devices, and ensure that the graph is consistent with the actual device operating status. For large-scale feedback, the system uses a distributed computing framework for efficient processing and online updates to ensure timely response and optimization. Through this feedback mechanism, the system can continuously optimize itself, improve recognition accuracy, and ensure that device recognition is always efficient and accurate in a constantly changing environment.

[0161] Network scanning tools, such as Nmap and Zmap, can efficiently detect device information in the network, including open ports, service types, and protocols, and quickly provide preliminary data on the device. At the same time, knowledge graphs, as a semantic network, can integrate information from different sources and establish multi-dimensional associations between devices and their features, functions, locations, etc. This combination can not only achieve real-time recognition, but also perform intelligent reasoning based on contextual information and device behavior patterns to improve recognition accuracy.

[0162] This embodiment can dynamically update device information by constructing a lightweight knowledge graph, supporting efficient identification and classification of a large number of IoT devices. This method will provide strong support for network security management, helping enterprises and organizations to quickly respond to potential threats and optimize resource allocation in an increasingly complex network environment. Therefore, the IoT device identification method based on lightweight knowledge graph has important practical significance and broad application prospects.

[0163] Dynamic graph construction + data sharding and on-demand loading:

[0164] The dynamic adjustment method based on real-time data stream has important application value in IoT device identification and knowledge graph construction. This method monitors the real-time data of the device and dynamically updates the structure and content of the knowledge graph, thereby improving the accuracy and flexibility of device identification. In the process of achieving lightweight, the storage and query efficiency is further optimized by combining data sharding and on-demand loading, making the entire system more efficient and intelligent.

[0165] First, the acquisition of real-time data streams is the basis for implementing this method. An efficient data stream acquisition mechanism needs to be established to collect device-related data in real time. This data includes sensor data, such as the real-time sensor output of the device (temperature, humidity, motion, etc.); user interaction data, which records the interaction between the user and the device (such as switch operation, adjustment settings, etc.); and network status information, which involves performance indicators such as the device's connection status, network latency, and bandwidth usage.

[0166] After data acquisition, real-time processing and analysis become key steps. By using the stream processing method that combines Kafka and Flink, specifically, first Kafka is responsible for collecting and transmitting the data generated by the device in real time and publishing it to different topics. Then, Flink, as a data processing framework, reads the data stream in Kafka in real time, performs feature extraction and event detection, and dynamically updates the structure and content of the knowledge graph. This combination not only ensures high throughput and reliability of data, but also realizes the real-time adaptability of the knowledge graph, and improves the accuracy and flexibility of IoT device identification. The feature extraction link will extract key features from real-time data, such as the frequency of use, failure rate, and working mode of the device, which provide the basis for subsequent analysis.

[0167] The core of this method is to implement a dynamic adjustment mechanism based on real-time analysis results. The structure of the knowledge graph will be dynamically updated according to the state changes and usage frequency of the device. For example, when a device is frequently used, the system can increase its importance in the graph and increase its association with other devices. At the same time, based on real-time data judgment, the addition and deletion of nodes and edges will also be implemented. If a device has not been used for a period of time, you can consider lowering its priority in the graph or temporarily removing it.

[0168] Optimization method of data sharding and on-demand loading: After initially building the knowledge graph and dynamically updating it, combine data sharding and on-demand loading methods to further optimize storage and query efficiency. With the increase in the number of devices and the real-time growth of device behavior data, the amount of graph data will become huge. At this time, data sharding and on-demand loading become key means to improve system performance.

[0169] Device information (such as the various states and features of IoT devices) can be stored in shards based on factors such as device type, manufacturer, and physical location, thereby avoiding loading all the information of all devices at once. When querying, the system only loads the subgraph part related to the current task, reducing the loading of redundant data. This on-demand loading method not only optimizes the consumption of storage and computing resources, but also speeds up the query speed and improves the response efficiency of the system.

[0170] During the dynamic update of the real-time data stream, the system only loads the data of the currently active device, and the inactive device information is loaded lazily through the on-demand loading mechanism. This ensures the real-time and flexibility of the graph, while greatly improving the performance of the system. For example, if a device has not been interacted with or used recently, the system may not load the detailed information of the device immediately, and only load the relevant data on demand when the device becomes an active device.

[0171] Dynamic update and optimization mechanism: In order to continuously optimize the dynamic adjustment process, we have also established a feedback mechanism. In actual IoT device recognition scenarios, user feedback can be collected through dedicated applications or interfaces. For example, users can rate the accuracy of device recognition when operating the device and provide specific opinions or suggestions. These feedbacks not only cover the recognition results of the device, but also the functional performance of the device, the relevance of the recommended content, etc. The system will use machine learning algorithms to analyze the collected user feedback to identify potential patterns and problems. For example, for certain device types, users may frequently report recognition errors. The system can focus on these feedbacks for in-depth analysis and extract features to improve the recognition algorithm. Through reinforcement learning algorithms, the system can adjust the dynamic update strategy based on historical feedback to better adapt to the actual needs of users.

[0172] In addition, the feedback mechanism can also collect users' experience and suggestions during use by periodically visiting them, ensuring the timeliness and representativeness of feedback information. For example, users can directly provide feedback on the recognition accuracy of the device in a specific environment during use, and the system will incorporate these feedbacks into the adjustment of the knowledge graph, thereby optimizing the device recognition strategy.

[0173] This feedback mechanism based on actual usage can not only enhance the system's responsiveness to dynamic changes, but also establish a good closed loop between recognition accuracy and user experience, ensuring that the IoT device recognition system continues to evolve to adapt to complex usage environments and user needs.

[0174] The method based on the dynamic adjustment of real-time data streams combined with data sharding and on-demand loading optimizes storage and query efficiency while ensuring timely updates of device information. Through data sharding and on-demand loading, the system avoids loading too much redundant data at one time, thereby improving the utilization of storage and computing resources and query response speed. At the same time, the knowledge graph can continuously make intelligent adjustments based on the dynamic behavior and real-time needs of the device. This optimization method not only improves the accuracy and response speed of IoT device identification, but also provides guarantees for the long-term stability and efficient operation of the system.

[0175] Embodiment 2

[0176] This embodiment provides an IoT device identification system based on a lightweight knowledge graph, including:

[0177] The preliminary identification module is configured to obtain data information of multiple dimensions and device types of IoT devices, calculate and sort the mutual information values ​​between the data information of multiple dimensions and the device types, select the data information of the corresponding dimensions with a mutual information value greater than a set threshold and fuse them with the device type to obtain a comprehensive feature;

[0178] The knowledge graph construction module is configured to construct the knowledge graph by taking the device type as the central node, taking each data of each dimension as a node, and taking the corresponding relationship between each node and the central node as an edge;

[0179] The knowledge graph matching module is configured to continue to identify data information of multiple dimensions and device types of IoT devices, determine the corresponding latest comprehensive features, calculate the similarity between the latest comprehensive features and the comprehensive features in the knowledge graph, and generate a candidate device type list based on the device types corresponding to the comprehensive features whose similarity is greater than the similarity threshold;

[0180] The context verification module is configured to use context information to verify the device type in the candidate device type list. If the verification is successful, the final matching device type is determined and the latest comprehensive features are updated to the corresponding device type in the knowledge graph; if the verification fails, the knowledge graph is updated based on the latest comprehensive features as a new node.

[0181] The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment 1. It should be noted that the above modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0182] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0183] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0184] Embodiment 3

[0185] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the method for identifying Internet of Things devices based on a lightweight knowledge graph as described in the first embodiment above are implemented.

[0186] Embodiment 4

[0187] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for identifying an Internet of Things device based on a lightweight knowledge graph as described in the first embodiment above are implemented.

[0188] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0189] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0190] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0191] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0192] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0193] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.< / locatedin> < / exhibitsbehavior> < / manufacturedby> < / http> < / supports> < / http> < / location> < / hasprotocol> < / type> < / manufacturer> < / location> < / http> < / hasprotocol> < / type>

Claims

1. The IoT device identification method based on lightweight knowledge graph is characterized by: include: Obtain data information and device types of multiple dimensions of IoT devices, calculate and sort the mutual information values ​​between the data information and device types of multiple dimensions, select the data information of the corresponding dimensions with mutual information values ​​greater than a set threshold and fuse them with the device type to obtain comprehensive features; Take the device type as the central node, each data in each dimension as a node, and the corresponding relationship between each node and the central node as an edge to build a knowledge graph; Continue to identify data information and device types of multiple dimensions of IoT devices, determine the corresponding latest comprehensive features, and calculate the similarity between the latest comprehensive features and the comprehensive features in the knowledge graph, and generate a list of candidate device types based on the device types corresponding to the comprehensive features whose similarity is greater than the similarity threshold; The context information is used to verify the device type in the candidate device type list. If the verification is successful, the final matching device type is determined, and the latest comprehensive features are updated to the corresponding device type in the knowledge graph; if the verification fails, the knowledge graph is updated based on the latest comprehensive features as a new node.

2. The method for identifying IoT devices based on a lightweight knowledge graph as claimed in claim 1, characterized in that: The data information in the multiple dimensions includes device ID, device name, device manufacturer, device IP address, device communication protocol, device port status, device location, device working scenario, device installation time, device maintenance record and device behavior.

3. The method for identifying IoT devices based on a lightweight knowledge graph as claimed in claim 1, characterized in that: The use of context information to verify the device type in the candidate device type list is specifically as follows: According to the candidate device type list, obtain the context node related to the device type; The context node of the IoT device to be verified is verified for consistency with each type in the candidate device type list. If the verification is successful, the final matching device type is determined; otherwise, the knowledge graph is updated based on the latest comprehensive features as a new node.

4. The method for identifying IoT devices based on a lightweight knowledge graph as claimed in claim 3, characterized in that: The context nodes include device location nodes, device time nodes, device maintenance record nodes, and device behavior nodes.

5. The method for identifying IoT devices based on a lightweight knowledge graph as claimed in claim 3, characterized in that: The consistency verification of the context node of the IoT device to be verified with each type in the candidate device type list is specifically as follows: Condition 1: Whether the device location node of the IoT device to be verified matches each type in the candidate device type list; Condition 2: Whether the typical behavior in the device behavior node of the IoT device to be verified matches each type in the candidate device type list; Condition 3: If the device installation time node or device maintenance record node of the IoT device to be verified is obtained, then determine whether it matches each type in the candidate device type list; otherwise, the device installation time node and device maintenance record node are not verified; If at least one condition is met, the consistency verification succeeds.

6. The method for identifying IoT devices based on a lightweight knowledge graph as claimed in claim 3, characterized in that: If there are multiple candidate device types that are successfully verified, the device type with the highest similarity is selected as the final matching device type.

7. The method for identifying IoT devices based on a lightweight knowledge graph as claimed in claim 1, characterized in that: If the similarity between the latest comprehensive feature and the comprehensive feature in the knowledge graph is less than the similarity threshold, the knowledge graph is updated based on the latest comprehensive feature as a new node.

8. The IoT device identification system based on lightweight knowledge graph is characterized by: include: The preliminary identification module is configured to obtain data information of multiple dimensions and device types of IoT devices, calculate and sort the mutual information values ​​between the data information of multiple dimensions and the device types, select the data information of the corresponding dimensions with a mutual information value greater than a set threshold and fuse them with the device type to obtain a comprehensive feature; The knowledge graph construction module is configured to construct the knowledge graph by taking the device type as the central node, taking each data of each dimension as a node, and taking the corresponding relationship between each node and the central node as an edge; The knowledge graph matching module is configured to continue to identify data information of multiple dimensions and device types of IoT devices, determine the corresponding latest comprehensive features, calculate the similarity between the latest comprehensive features and the comprehensive features in the knowledge graph, and generate a candidate device type list based on the device types corresponding to the comprehensive features whose similarity is greater than the similarity threshold; The context verification module is configured to use context information to verify the device type in the candidate device type list. If the verification is successful, the final matching device type is determined and the latest comprehensive features are updated to the corresponding device type in the knowledge graph; if the verification fails, the knowledge graph is updated based on the latest comprehensive features as a new node.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the method for identifying an Internet of Things device based on a lightweight knowledge graph as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the method for identifying an Internet of Things device based on a lightweight knowledge graph as described in any one of claims 1 to 7 are implemented.

Images