Data access method and system of intelligent Internet of Things gateway

Abstract

The invention relates to a data access method and system of an intelligent Internet of Things gateway. The method comprises the following steps: performing feature matching and protocol identification on a plurality of sensors and terminal equipment connected to the intelligent Internet of Things gateway by using an equipment feature library to obtain an equipment protocol type; analyzing the original data to obtain standardized data; performing encryption processing to obtain an encrypted security data packet and an encrypted session key; performing multi-factor transmission channel selection to obtain an optimal transmission channel combination; and based on the optimal transmission channel combination, sending the encrypted security data packet to the Internet of Things platform, verifying the received data confirmation information based on a consensus algorithm, and outputting a data access state report through a data visual interface and an AI voice assistant. The processing strategy can be dynamically adjusted according to the equipment characteristics, the data characteristics and the network environment, and the flexibility and adaptability of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of smart Internet of Things gateways, and in particular to a data access method and system for a smart Internet of Things gateway. Background Art

[0002] As a key node connecting various sensors, terminal devices and cloud platforms, smart IoT gateways play an increasingly important role in data collection, processing and transmission. However, the current IoT devices are diverse and have different communication protocols, which poses a huge challenge to device access and data standardization. Traditional IoT gateways are often inefficient in processing diverse devices and protocols, and are unable to meet the access needs of large-scale heterogeneous devices.

[0003] In addition, the security and privacy protection of IoT data are also receiving increasing attention. In the process of data transmission, how to effectively protect sensitive information and prevent data leakage and tampering has become an urgent problem to be solved. At the same time, the complex and changing network environment has also brought challenges to the reliable transmission of data. How to achieve efficient and stable data transmission under limited network resources has become the key point of IoT system optimization. Summary of the invention

[0004] The main purpose of the present invention is to provide a data access method and system for a smart Internet of Things gateway. The present invention can dynamically adjust the processing strategy according to device characteristics, data characteristics and network environment, thereby improving the flexibility and adaptability of the system.

[0005] To achieve the above object, the present invention provides a data access method for a smart Internet of Things gateway, comprising the following steps:

[0006] Use the device feature library to perform feature matching and protocol identification on multiple sensors and terminal devices connected to the smart IoT gateway to obtain the device protocol type;

[0007] Activate the corresponding protocol parsing plug-in based on the device protocol type, parse the original data to obtain the parsing result, and call the edge computing microservice to input the parsing result and the original data into the semantic ontology model for data standardization processing to obtain standardized data;

[0008] Calculating the sensitivity of the standardized data, and selecting an encryption algorithm according to the sensitivity to encrypt the standardized data to obtain an encrypted security data packet and an encrypted session key;

[0009] Perform a multi-dimensional evaluation on the current network environment of the smart IoT gateway to obtain network evaluation data, and input the network evaluation data into a deep reinforcement learning model to perform multi-factor transmission channel selection to obtain an optimal transmission channel combination;

[0010] Based on the optimal transmission channel combination, the encrypted security data packet is sent to the IoT platform, the received data confirmation information is verified based on the consensus algorithm, and the data access status report is output through the data visualization interface and AI voice assistant.

[0011] The present invention also provides a data access system for a smart Internet of Things gateway, comprising:

[0012] An identification unit is used to perform feature matching and protocol identification on multiple sensors and terminal devices connected to the smart IoT gateway using a device feature library to obtain a device protocol type;

[0013] A parsing unit, configured to activate a corresponding protocol parsing plug-in based on the device protocol type, parse the original data to obtain a parsing result, and call an edge computing microservice to input the parsing result and the original data into a semantic ontology model for data standardization processing to obtain standardized data;

[0014] An encryption unit, used to calculate the sensitivity of the standardized data, and select an encryption algorithm according to the sensitivity to encrypt the standardized data to obtain an encrypted security data packet and an encrypted session key;

[0015] A selection unit, configured to perform a multi-dimensional evaluation on the current network environment of the smart IoT gateway to obtain network evaluation data, and input the network evaluation data into a deep reinforcement learning model to perform multi-factor transmission channel selection to obtain an optimal transmission channel combination;

[0016] The output unit is used to send the encrypted security data packet to the IoT platform based on the optimal transmission channel combination, verify the received data confirmation information based on the consensus algorithm, and output the data access status report through the data visualization interface and AI voice assistant.

[0017] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0018] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0019] In summary, the technical solution provided by the present invention realizes automatic identification and protocol matching of various sensors and terminal devices through device feature library and protocol identification technology, improves the efficiency and accuracy of device access, and solves the problem of heterogeneous device access. By using semantic ontology model and edge computing technology, the original data is intelligently parsed and standardized, and the unified expression of data from different sources is realized. Through sensitivity assessment and dynamic encryption algorithm selection, the data is differentially encrypted, which not only ensures data security, but also optimizes the efficiency of encryption resources. The deep reinforcement learning model is used for multi-factor transmission channel selection, which realizes the intelligent scheduling and optimal utilization of network resources, and improves the efficiency and reliability of data transmission. Through the data visualization interface and AI voice assistant, the intuitive display and voice broadcast of data access status are realized, which improves the convenience of system operation and maintenance and user experience. The present invention has strong adaptive ability, can dynamically adjust the processing strategy according to device characteristics, data characteristics and network environment, and improves the flexibility and adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a schematic diagram of the steps of a data access method for a smart Internet of Things gateway in one embodiment of the present invention;

[0021] Figure 2 This is a structural block diagram of a data access system of a smart Internet of Things gateway in one embodiment of the present invention;

[0022] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0023] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0025] Reference Figure 1 This embodiment provides a data access method for a smart Internet of Things gateway, including the following steps:

[0026] S1, using the device feature library to perform feature matching and protocol identification on multiple sensors and terminal devices connected to the smart IoT gateway to obtain the device protocol type;

[0027] Among them, the communication interface of the smart IoT gateway is monitored in real time to identify newly connected sensors and terminal devices. The physical connection features of sensors and terminal devices are recorded to generate a device connection information list. The physical connection features include the hardware identifier, MAC address, connection port, etc. of the device. The protocol identifiers, data frame formats, and communication parameters of multiple pre-stored communication protocols are extracted from the device feature library, and a protocol feature matrix is ​​constructed based on these features. The protocol identifier describes the characteristics of a specific protocol, such as the communication type, data packet structure, and protocol version, while the data frame format and communication parameters include the frame header, frame trailer, and related checksum fields in the protocol communication. After constructing the protocol feature matrix, the physical connection features in the device connection information list are calculated for similarity with the protocol feature matrix to obtain the similarity score between the device and each protocol. Based on the similarity score, several candidate protocols with the highest similarity are selected for each device to form a candidate protocol set for the device. For each communication protocol in the device candidate protocol set, packet capture and protocol simulation are performed. The matching degree of each protocol is calculated by sending simulated data packets to the device and analyzing the device's response. By simulating a real communication environment, detecting the response of the device to a specific protocol, evaluating the compatibility of the device and the protocol, and generating a protocol matching score table. The protocol matching score table contains the matching score of each candidate protocol. Combined with the previous similarity score, the two indicators are comprehensively considered to select the protocol with the highest votes for each device as the final identification result, and obtain a preliminary mapping relationship between the device and the protocol. In the preliminary device-protocol mapping relationship, adaptive protocol learning is performed for devices with a confidence level lower than the preset threshold. The key features of these devices in the communication process are extracted, and the key features are compared with the known protocols in the device feature library to generate a temporary protocol. The temporary protocol is applied to the real-time data stream of the corresponding device for protocol parsing test. The effectiveness of the temporary protocol is evaluated by testing and calculating the parsing success rate and data integrity index. If the parsing success rate and data integrity index of the temporary protocol meet the preset conditions, it means that the temporary protocol can parse the device data stably and accurately and meet the access requirements. Therefore, the temporary protocol is added to the device feature library, and the device-protocol mapping relationship is updated accordingly to obtain the protocol type of the device. By continuously expanding the device feature library, the system's protocol recognition capability and compatibility will gradually increase, enabling it to adapt to a wider range of device types and communication protocols.

[0028] S2, activates the corresponding protocol parsing plug-in based on the device protocol type, parses the original data, obtains the parsing result, and calls the edge computing microservice to input the parsing result and the original data into the semantic ontology model for data standardization to obtain standardized data;

[0029] Specifically, according to the device protocol type, a matching protocol parsing plug-in is retrieved from the protocol parsing plug-in library. The protocol parsing plug-in is pre-defined with parsing rules for different protocols during design, which can effectively support protocol parsing of various devices. The protocol parsing plug-in is loaded into the parsing engine of the smart IoT gateway to ensure that the plug-in can run normally and provide parsing services in subsequent data processing links. After the protocol parsing plug-in is successfully loaded, the original data is input into the plug-in, and the original data is decomposed, extracted and converted based on the preset parsing rules to obtain the corresponding parsing results. The parsing rules include operations such as segmentation of data frames, extraction of specific fields, and conversion of data formats, so as to convert data from the original protocol format of a specific device into a more general form. The parsing results are checked for data integrity, including calculating the data checksum, comparing the checksum with the preset threshold, generating a data validity mark, and identifying whether the data is abnormal or damaged during transmission and parsing. The parsing results and data validity marks are transmitted to the edge computing microservice, and the edge computing task is triggered at the same time. The edge computing microservice processes the data on the local gateway, such as format conversion, preliminary statistical analysis or filtering, to generate preliminarily processed data. The semantic ontology model related to the device protocol type is extracted from the semantic ontology library to build a semantic mapping rule set. The semantic ontology model is a knowledge-based description of device data and device attributes. These models are used to convert the original data of different devices into a unified semantic expression. The semantic mapping rule set is a mapping rule built based on the semantic ontology model, which guides how to semantically annotate the preliminarily processed data and the original data. Using these rules, after annotating the data, a data entity with semantic tags is generated, which contains the basic content of the device data and also integrates the semantic information about the device and its data, making the meaning of the data clearer and facilitating subsequent reasoning and analysis. Ontology reasoning is performed on the data entity with semantic tags. By using the description logic algorithm to analyze and deduce the relationship between the data, the description logic algorithm can reveal the implicit relationship between the data, and logically associate different data entities to build a semantic association network. Based on the constructed semantic association network, the data is normalized to obtain standardized data.

[0030] Furthermore, according to the device protocol type, a query statement containing a protocol identifier and a keyword is constructed, and the semantic ontology library is searched using the SPARQL query language to extract multiple candidate ontology models related to the device protocol type. SPARQL is a query language used to query the ontology library. Through precise query conditions, the ontology model related to the device protocol type can be effectively found. Each candidate ontology model is structurally analyzed, and the core concepts, attributes and relationship information of each ontology model are extracted, and the ontology feature vector is constructed based on this. The ontology feature vector is used to describe the relationship between concepts in the ontology model and its attribute characteristics. Each feature is given a weight to form a weighted ontology feature vector set. The feature weight is calculated based on its importance in the ontology structure. For example, the core concepts and key relationships have higher weights, while the secondary attributes have lower weights. The weighted feature vector set helps to quantify the overall characteristics of the ontology model. The weighted ontology feature vector set is matched with the knowledge graph of the corresponding domain for similarity, and the candidate ontology model with the highest similarity is selected as the most matching semantic ontology model. As rich background knowledge, the domain knowledge graph effectively supports the precise matching of the ontology model through its structure and semantic information. By matching the similarity with the knowledge graph, the ontology model with the highest relevance to the device protocol type is selected to ensure that the selected ontology model has high semantic relevance. Based on the selected semantic ontology model, the class, attribute and relationship information defined therein are extracted to construct a semantic mapping rule set. The semantic mapping rule set is used to describe the mapping relationship between data fields and semantic concepts, as well as the attribute constraints and inference rules between these concepts. The preliminarily processed data and the original data are parsed into key-value pairs, and the corresponding semantic concepts and attributes are assigned to each key-value pair according to the semantic mapping rule set to generate preliminary semantic annotation results. The key-value pair parsing process extracts each field in the data and matches it with the concepts and attributes in the semantic mapping rule set to give the data field clear semantic labels and attributes, thereby realizing the preliminary semanticization of the data. The preliminary semantic annotation results are contextually analyzed, and the association information between adjacent data items is extracted. The mutual information value between each data item is calculated to construct the data item association matrix. The mutual information value reflects the degree of association between data items, and the data item association matrix describes the connection between different data items in a semantic and statistical sense. By constructing the association matrix, the deep semantic relationship between data items is mined. On this basis, based on the data item association matrix, the vector representation of the semantic concept in the knowledge graph is calculated, and the semantic concept with the highest cosine similarity is selected as the final semantic tag through the calculation of cosine similarity. According to the final semantic tag obtained, the preliminary semantic annotation results are converted into the Resource Description Framework (RDF) format. RDF format is a standard format for describing resources and their relationships. Each data item is represented as a subject-predicate-object triple in RDF, thereby constructing the association relationship between data.A globally unique URI identifier is assigned to each triple to ensure that the data entity is unique and identifiable globally, ultimately obtaining a data entity with a semantic tag.

[0031] S3, calculating the sensitivity of the standardized data, and selecting an encryption algorithm according to the sensitivity to encrypt the standardized data to obtain an encrypted security data packet and an encrypted session key;

[0032] It should be noted that the feature extraction of the standardized data is performed, and the data field is mapped to a vector representation using a pre-trained word embedding model to obtain a data feature vector set. The word embedding model is a model pre-trained by natural language processing technology, which converts the information in the data field into a numerical vector form with a certain semantic representation, so that the characteristics of the data field can be effectively expressed in the vector space. The data feature vector set is input into the sensitivity assessment model for sensitivity assessment. The sensitivity assessment model is based on the multi-layer perceptron structure. The multi-layer perceptron is a feedforward neural network with strong feature learning ability and is suitable for processing high-dimensional numerical data. During the model training process, the cross entropy loss function is used to ensure that the model can accurately distinguish data fields with different sensitivities. The model outputs the sensitivity score of each data field through a deep analysis of the data features to obtain a sensitivity score set. The sensitivity score set is clustered. The data fields are divided into different sensitivity levels to form a data sensitivity classification result. Through the clustering method, data fields with similar sensitivity scores are clustered together to form several sensitivity levels, such as low sensitivity, medium sensitivity and high sensitivity levels. According to the data sensitivity classification results, the corresponding encryption algorithm is selected from the predefined encryption algorithm library. For low-sensitivity data fields, a lightweight encryption algorithm is selected. For high-sensitivity data fields, a strong encryption algorithm, such as AES (Advanced Encryption Standard), is used to ensure the security of data during transmission and storage, and an encryption algorithm selection scheme is obtained. Based on the encryption algorithm selection scheme, data fields with different sensitivity levels are encrypted respectively, and encryption session keys are generated at the same time to obtain preliminary encrypted data and encryption session keys. In this process, each data field selects a different encryption algorithm for processing according to its sensitivity level, and the encryption session key is used for verification when decrypting data. The generated encryption session key is unique for each encrypted data segment to ensure the security and non-tamperability of the data during transmission. After the preliminary encryption is completed, the encrypted data is processed in blocks, and a unique identifier and version information are added to each data block to construct an encrypted data packet structure to obtain structured encrypted data. Data block processing is to improve the efficiency of the transmission process and the flexibility of data management. The unique identifier and version information of each data block ensure that the data can be correctly reassembled and decoded at the receiving end, and the version of the data can be traced, thereby ensuring the consistency of the data. On this basis, integrity protection is performed on structured encrypted data, the HMAC (key-based message authentication code) value of each data packet is calculated and attached to the data packet to obtain an encrypted data packet with integrity protection. The calculation of the HMAC value is to ensure that the data has not been tampered with during transmission. The receiver verifies the HMAC value to determine the integrity of the data packet to prevent malicious attacks and data leakage.The encrypted session key is encrypted using an asymmetric encryption algorithm, and the encrypted session key is combined with the integrity-protected encrypted data packet to obtain an encrypted security data packet and an encrypted session key. Asymmetric encryption algorithms such as RSA are used to encrypt the session key. This encryption method has the characteristics of public key encryption and private key decryption, which can ensure the security of the encrypted session key during transmission.

[0033] S4, conduct a multi-dimensional evaluation of the current network environment of the smart IoT gateway to obtain network evaluation data, and input the network evaluation data into the deep reinforcement learning model to perform multi-factor transmission channel selection to obtain the optimal transmission channel combination;

[0034] Specifically, all network interfaces of the smart IoT gateway are scanned in parallel, and the status information of the interfaces is collected through SNMP (Simple Network Management Protocol) to obtain a list of network interfaces. The SNMP protocol is a protocol used for network device management, through which important information such as the working status and traffic statistics of each network interface can be efficiently obtained. By monitoring the status of the network interface in real time, it is determined which interfaces are active and which interfaces are available. Based on the network interface list, a network performance test is performed on each active interface to evaluate the quality of each network interface. The round-trip delay of each interface is measured using the ICMP protocol, the bandwidth of each interface is tested using the iPerf tool, the packet loss rate is measured using the UDP traffic generator, and the signal strength and signal-to-noise ratio of the wireless interface are additionally measured to obtain the original network performance data matrix containing multi-dimensional indicators. The ICMP protocol is used to test network connectivity and delay, and its round-trip delay can reflect the delay characteristics of network transmission; the iPerf tool is a commonly used bandwidth testing tool, which measures the actual bandwidth of each network interface by generating TCP or UDP traffic; the UDP traffic generator is used to measure the packet loss rate, reflecting the stability of the network interface under load. The signal strength and signal-to-noise ratio of the wireless interface are important indicators for measuring the quality of wireless networks. These test data are used to evaluate the performance of each network interface. A network interface evaluation vector is generated based on the original network performance data matrix. The comprehensive score of each interface is obtained by weighted fusion of multiple performance indicators. Based on the comprehensive score, the network interfaces are sorted in descending order, and the top N interfaces are selected as the candidate transmission channel set. The network performance data of the candidate transmission channel set is fused with the historical transmission data in time series to construct a state space matrix. Each row of the state space matrix represents the network state at a time point, and the columns include network performance indicators such as bandwidth utilization and delay jitter. By fusion of time series, the information in the historical data is fully utilized to capture the dynamic change characteristics of network performance and describe the current network state. The current network state representation is input into the deep reinforcement learning model. The deep reinforcement learning model adopts a dual DQN structure, which is an improved Q learning algorithm. The dual DQN structure uses two independent Q networks to reduce the problem of overestimation of Q values, thereby improving the stability of the model and the accuracy of decision-making. In the forward propagation process of the model, the Q value of each possible transmission channel combination is calculated to obtain a transmission channel evaluation matrix containing all possible combination Q values. The Q value indicates the quality of each transmission channel combination under the current network status. The higher the Q value, the better the performance of the transmission channel combination and the more suitable it is for the current data transmission needs. Based on the transmission channel evaluation matrix, the probability of each transmission channel combination being selected is calculated. By performing probability distribution calculations on all possible channel combinations, the priority of each combination in the current situation is quantified, and the combination with the highest probability is selected as the optimal transmission channel combination.

[0035] S5, based on the optimal transmission channel combination, sends the encrypted security data packet to the IoT platform, verifies the received data confirmation information based on the consensus algorithm, and outputs the data access status report through the data visualization interface and AI voice assistant.

[0036] Among them, according to the optimal transmission channel combination, the encrypted security data packet is fragmented, and the large data packet is divided into multiple small data packets. Each data fragment is attached with a unique sequence number and checksum to ensure the order and integrity of the fragment, and a data fragment set is obtained. The purpose of fragmentation is to improve the flexibility and efficiency of data transmission, so that each fragment can be transmitted independently through different transmission paths, and the available bandwidth and transmission channel resources are effectively utilized. Based on the data fragment set, the optimal transmission path of each fragment is calculated. Combined with the optimal transmission channel combination, the transmission path selection is performed for each data fragment, and different fragments are allocated to different transmission channels to form a fragment transmission scheme, so as to ensure that the transmission process of each fragment does not interfere with each other, maximize the use of the bandwidth and resources of the transmission channel, and improve the overall data transmission performance and stability. Through the dynamic fragment allocation strategy, the transmission path of each fragment is flexibly adjusted according to the changes in the network environment, so as to achieve efficient transmission in a complex network environment. According to the fragment transmission scheme, multi-threaded parallel transmission technology is used to send data fragments to the IoT platform at the same time. In the process of parallel transmission, in order to ensure the reliability of data, the timeout retransmission mechanism is started, that is, whenever a data slice is detected to be not successfully delivered within the predetermined time, the retransmission operation is triggered to ensure that the data can finally reach the target platform. At the same time, the sending status of each data slice is recorded, including success, failure, number of retransmissions and other information, to form a data transmission log. When the IoT platform successfully receives the data, a data confirmation message is returned. The data confirmation message contains the serial number of the successfully received data slice and a digital signature to ensure the authenticity and integrity of the data. The digital signature is verified using an asymmetric encryption algorithm to ensure that the data is indeed sent by the IoT platform and has not been tampered with during the transmission process. In this way, the validity of the data is preliminarily verified to ensure the security of the data during transmission. The preliminary verification results are input into the consensus algorithm model based on Byzantine fault tolerance. The Byzantine fault tolerance consensus algorithm is an algorithm used to solve the consistency problem between nodes in a distributed system. The validity of the data is verified by performing multiple rounds of voting among multiple distributed nodes. When more than 2 / 3 of the nodes reach a consensus on a data slice, it is considered that the data slice has been correctly received, and the confirmation result of data reception is output. Based on the data reception confirmation results, the data transmission success rate, average delay and throughput are calculated to construct a data access status report. The data transmission success rate reflects the proportion of data packets that successfully reach the IoT platform during the transmission process, while the average delay is used to measure the average time required for data to be sent to confirmation, and the throughput represents the amount of data successfully transmitted per unit time. Through the comprehensive calculation of these indicators, the data access performance of the smart IoT gateway is comprehensively evaluated and a data access status report is generated. Through the data visualization interface, the content in the data access status report is displayed in the form of interactive charts.The AI ​​voice assistant uses text-to-speech (TTS) technology to convert data access status reports into voice output, allowing users to obtain data transmission status information through voice.

[0037] In one example, a device feature library is used to perform feature matching and protocol identification on multiple sensors and terminal devices connected to a smart IoT gateway to obtain the device protocol type, including: real-time monitoring of the communication interface of the smart IoT gateway, identifying newly connected sensors and terminal devices, and recording the physical connection features of the sensors and terminal devices to obtain a device connection information list; extracting the protocol identifiers, data frame formats, and communication parameters of multiple pre-stored communication protocols from the device feature library to construct a protocol feature matrix; calculating the similarity between the physical connection features in the device connection information list and the protocol feature matrix to obtain a device-protocol similarity score, and based on the device-protocol similarity score, selecting the K candidate protocols with the highest similarity for each device to obtain a device candidate protocol set; for each communication protocol in the device candidate protocol set, , perform data packet capture and protocol simulation, calculate the protocol matching degree by sending simulated data packets to the device and analyzing the response, and obtain the protocol matching score table; according to the protocol matching score table, comprehensively consider the similarity score and the protocol matching degree, select the protocol with the highest votes as the device identification result, and obtain the preliminary device-protocol mapping relationship; for the devices with confidence levels lower than the preset threshold in the preliminary device-protocol mapping relationship, perform adaptive protocol learning and extract key features, compare the key features with the known protocol library, and generate a temporary protocol; apply the temporary protocol to the real-time data stream of the corresponding device, perform protocol parsing test, calculate the parsing success rate and data integrity index, and if the parsing success rate and data integrity index meet the preset conditions, add the temporary protocol to the device feature library, and update the device-protocol mapping relationship to obtain the device protocol type.

[0038] In this example, the communication interface of the smart IoT gateway is monitored in real time through periodic polling or interrupt triggering. When a new device is connected, the system can quickly detect the connection of the device and record the physical connection characteristics of the device, including the device's MAC address, port number, device type and other information, to form a device connection information list. The protocol identifiers, data frame formats, and communication parameters of the pre-stored multiple communication protocols are extracted from the device feature library, and a protocol feature matrix is ​​constructed. The protocol feature matrix is ​​used to describe the characteristics of each protocol. The protocol feature matrix is ​​expressed as:

[0039] P = {p 1 ,p 2 ,…,p m};

[0040] Where P represents the protocol feature matrix, p iThe feature vector representing protocol i includes the protocol identifier, data frame format, communication parameters, etc. The physical connection features in the device connection information list are similar to the protocol feature matrix to obtain the similarity score between the device and the protocol. The similarity is calculated using the cosine similarity method, and the calculation formula is as follows:

[0041]

[0042] Among them, S ij represents the similarity score between device i and protocol j, V i represents the physical connection feature vector of device i, p j represents the feature vector of protocol j, the symbol · represents the dot product of two vectors, ||V i || and ||p j|| represents the modulus of the device feature vector and the protocol feature vector, respectively. Based on the similarity score, the K candidate protocols with the highest similarity are selected for each device to obtain the candidate protocol set of the device. For example, suppose there is a newly connected temperature sensor, and the system detects that its physical characteristics include a specific MAC address, the port number of the connection, and the initial handshake packet type. These features are similarly calculated with multiple protocols in the protocol feature library. Assuming that high similarity scores are obtained for several protocol types, such as Modbus protocol, MQTT protocol, etc., these K candidate protocols are used as the protocol types supported by the device. In order to determine the specific communication protocol, packet capture and protocol simulation are performed for each communication protocol in the candidate protocol set. The protocol matching degree is calculated by sending simulated data packets to the device and analyzing the device's response to these data packets. By capturing the communication data packets between the device and the IoT gateway, simulating the standard operations of each candidate protocol, observing whether the device can respond correctly, and obtaining the protocol matching score table. Assuming that the Modbus protocol is simulated and a specific read register request is sent, if the response returned by the device conforms to the standard data frame format of the Modbus protocol, it means that the device has a high probability of supporting the Modbus protocol. According to the protocol matching score table, the similarity scores and protocol matching between the device and each protocol are comprehensively considered, and the protocol with the highest votes is selected as the final identification result of the device to obtain a preliminary device-protocol mapping relationship. For devices with confidence levels lower than the preset threshold, adaptive protocol learning is performed. The process of adaptive protocol learning includes in-depth analysis of the device's data stream and extraction of key communication features. These key features are compared with the known protocol library. If some features cannot be matched to the known protocol, a temporary protocol is generated to describe these new features. For example, the communication protocol of some new custom devices has not been widely standardized. By learning their data interaction behavior and extracting the feature fields in the frame format, a temporary protocol is generated. The generated temporary protocol is applied to the real-time data stream of the device and a protocol parsing test is performed to verify the validity of the protocol. The success rate and data integrity index of the temporary protocol parsing are calculated. If the parsing success rate and data integrity index meet the preset conditions, it means that the temporary protocol can correctly describe the communication behavior of the device. The temporary protocol is added to the device feature library, and the device-protocol mapping relationship is updated to finally obtain the protocol type of the device.

[0043] In one example, the corresponding protocol parsing plug-in is activated based on the device protocol type, the original data is parsed to obtain the parsing result, and the edge computing microservice is called to input the parsing result and the original data into the semantic ontology model for data standardization processing to obtain standardized data, including: according to the device protocol type, the matching protocol parsing plug-in is retrieved from the protocol parsing plug-in library, and the protocol parsing plug-in is loaded into the parsing engine of the smart IoT gateway to obtain the protocol parsing plug-in; the original data is input into the protocol parsing plug-in, and the original data is decomposed, extracted and converted using the preset parsing rules to obtain the parsing result, and the data integrity check is performed on the parsing result to calculate the data Checksum and compare it with the preset threshold to obtain the data validity mark; transmit the parsing result and data validity mark to the edge computing microservice, and trigger the edge computing task to obtain the preliminary processed data; extract the semantic ontology model related to the device protocol type from the semantic ontology library, build a semantic mapping rule set, and semantically annotate the preliminary processed data and the original data according to the semantic mapping rule set to generate data entities with semantic tags; perform ontology reasoning on the data entities with semantic tags, use description logic algorithms to analyze and deduce data relationships, obtain a semantic association network, and perform normalization based on the semantic association network to obtain standardized data.

[0044] In this example, according to the device protocol type, a query is performed in the protocol parsing plug-in library to find the parsing plug-in corresponding to the protocol used by the device. Assuming that the protocol type of a device is Modbus protocol, the Modbus protocol parsing plug-in is extracted from the plug-in library and loaded into the parsing engine of the smart IoT gateway. After loading, the original data is input into the protocol parsing plug-in, and the original data is decomposed, extracted and converted using the parsing rules preset in the plug-in to obtain a structured parsing result. The parsing rules in the protocol parsing plug-in are defined as a specific data frame structure and a method for extracting fields. For example, the data frame of the Modbus protocol usually consists of a function code, a start address, a data length, and a data value. The parsing plug-in extracts the original data in segments according to these rules, decomposes each field and performs the corresponding data type conversion to obtain the parsed result. Assuming that the data sent by the device is a read register request, the parsing plug-in extracts the function code, address, and data value respectively, and converts these fields into structured data that is easy to process. The parsing result is checked for integrity to ensure that the data has not been tampered with during transmission and parsing. The integrity check process is completed by calculating the checksum of the data. The calculation formula for the checksum is:

[0045]

[0046] Among them, C represents the calculated data checksum, D irepresents the parsed data field, n is the total number of data fields, and M is a preset constant for modulo operation. A checksum value is obtained through this formula, and then the value is compared with the preset threshold to determine the validity of the data. If the calculated checksum value matches the preset checksum value, the data is marked as valid, otherwise the data is damaged or tampered with during transmission. The parsed result and data validity mark are transmitted to the edge computing microservice to trigger the corresponding edge computing task. The edge computing microservice can perform preliminary calculation and processing on the data, reduce the processing pressure of the central system, and improve the efficiency of data processing. After completing the edge computing task, the data after preliminary processing is semantically annotated to improve the understandability and operability of the data in the Internet of Things platform. The semantic ontology model related to the device protocol type is extracted from the semantic ontology library to construct a semantic mapping rule set. The semantic ontology model is a knowledge representation method that defines concepts and their relationships in a specific field. For example, the ontology model corresponding to the Modbus protocol contains concepts such as "function code" and "register address". Based on these concepts, a semantic mapping rule set is constructed, which defines the mapping relationship between data fields and semantic concepts. The preliminarily processed data and the original data are semantically annotated according to the semantic mapping rule set to generate data entities with semantic tags. Ontology reasoning is performed on the data entities with semantic tags, and the hidden associations in the data are revealed by analyzing and deducing the relationships between the data. The reasoning process relies on description logic algorithms. Description logic is a logical tool for knowledge representation and reasoning that can effectively handle the hierarchical structure and attribute relationships between concepts. Assuming there is data from a temperature sensor, reasoning reveals the potential association between the data "high temperature" and "device abnormality alarm", and a semantic association network is constructed. The formula for description logic reasoning is as follows:

[0047] R=f(C 1 ,C 2 ,…,C m );

[0048] Among them, R represents the inferred relationship, C 1 ,C 2 ,…,C m represents the concept set involved in reasoning, and f is the reasoning function. The relationship between concepts is revealed through the reasoning function to form a semantic association network. Based on the semantic association network, the data is normalized to obtain standardized data. Data from different devices and different protocols are converted into a unified format so that they can be uniformly processed and analyzed in the IoT platform.

[0049] In one example, a semantic ontology model related to the device protocol type is extracted from a semantic ontology library, a semantic mapping rule set is constructed, and the preliminarily processed data and the original data are semantically annotated according to the semantic mapping rule set to generate a data entity with a semantic tag, including: constructing a query statement according to the device protocol type, the query statement includes a protocol identifier and a keyword, searching the semantic ontology library using the SPARQL query language, extracting the ontology model related to the device protocol type, and obtaining a plurality of candidate ontology models; performing structural analysis on each candidate ontology model, extracting core concepts, attributes, and relationships, constructing an ontology feature vector, and calculating feature weights to obtain a weighted ontology feature vector set; performing similarity matching between the weighted ontology feature vector set and the corresponding domain knowledge graph, and selecting the candidate ontology model with the highest similarity as the most matching semantic ontology model; extracting classes, attributes, and relationships based on the semantic ontology model; and relationship definitions, and construct a semantic mapping rule set, which includes mapping relationships from data fields to semantic concepts, attribute constraints, and inference rules; parse the preliminarily processed data and raw data into key-value pairs, and assign corresponding semantic concepts and attributes to each key-value pair according to the semantic mapping rule set to generate preliminary semantic annotation results; perform context analysis on the preliminary semantic annotation results, extract association information between adjacent data items, calculate mutual information values, and construct a data item association matrix; at the same time, based on the data item association matrix, calculate the vector representation of the semantic concept in the knowledge graph, and select the semantic concept with the highest cosine similarity as the final semantic tag; convert the preliminary semantic annotation results into a resource description framework format according to the final semantic tag, each data item is represented as a subject-predicate-object triple, and each triple is assigned a globally unique URI identifier to obtain a data entity with a semantic tag.

[0050] In this example, a query statement is constructed according to the device protocol type. The query statement contains a protocol identifier and keywords to describe the semantic features related to the protocol. For example, for a device using the Modbus protocol, the query statement may contain the protocol identifier "Modbus" and keywords such as "register" and "function code". Through these keywords, the semantic requirements related to the device protocol type are described. The semantic ontology library is retrieved using the SPARQL query language. SPARQL is a query language specifically used to query semantic networks and ontology libraries, similar to the role of SQL in relational databases. Through the constructed query statement, multiple candidate ontology models related to the device protocol type are extracted from the semantic ontology library. For example, in the query process of the Modbus protocol, the SPARQL query statement matches relevant ontology models such as "Modbus communication model" and "industrial protocol model" to obtain a set of candidate ontology models. Each candidate ontology model is structurally analyzed to extract core concepts, attributes and relationships from it to construct an ontology feature vector. The feature vector is used to quantitatively represent the structure and semantic features of the ontology model. The feature vector is expressed as:

[0051] F i ={c 1 ,c 2 ,…,c n ,r 1 ,r 2 ,…,r m};

[0052] Among them, F i represents the feature vector of candidate ontology model i, c 1 ,c 2 ,…,c n Represents the core concept in the ontology model, r 1 ,r 2 ,…,r m Represents the relationship between these concepts. Through in-depth analysis of the ontology model structure, the key concepts and attribute relationships in the ontology model are extracted to form a complete feature description. The feature weight is calculated for each feature to obtain a weighted ontology feature vector set. The feature weight is calculated based on the frequency of occurrence of concepts and relationships in the ontology model and their importance. For example, a higher weight is given to core concepts that appear frequently in the model and are important for describing the model. The weighted ontology feature vector set is expressed as:

[0053] W i ={w 1 ×c 1 ,w 2 ×c 2 ,…,w n ×c n ,wn+1 × 1 ,…,w n+m × m};

[0054] Among them, W i is the weighted feature vector set of candidate ontology model i, w j Represents feature c j or j The weight of the weighted ontology feature vector set is matched with the corresponding domain knowledge graph for similarity. The knowledge graph contains various concepts in the field and their associations. By matching the weighted feature vector with the concept vector in the knowledge graph, the semantic ontology model closest to the device protocol type is found. The similarity matching adopts the cosine similarity method, and the calculation formula is as follows:

[0055]

[0056] Among them, Similarity represents the weighted feature vector W i The similarity with the concept vector K in the knowledge graph. The symbol · represents the dot product of two vectors, ||W i || and ||K|| represent the modulus of the weighted feature vector and the knowledge graph vector, respectively. Through similarity matching, the candidate ontology model with the highest similarity is selected as the most matching semantic ontology model. Based on the most matching semantic ontology model, class, attribute and relationship definitions are extracted to construct a semantic mapping rule set. The semantic mapping rule set is used to describe the mapping relationship between device data fields and semantic concepts, including attribute constraints and inference rules. For example, in the semantic ontology model of the Modbus protocol, "register address" is mapped to the semantic concept of "address", and "function code" is mapped to "operation type", etc. Through the semantic mapping rule set, the device data is associated with a specific semantic concept. The preliminarily processed data and the original data are parsed into key-value pairs. According to the semantic mapping rule set, the corresponding semantic concept and attribute are assigned to each key-value pair to generate a preliminary semantic annotation result. For example, for the temperature data collected from the Modbus device, after semantic mapping, the "temperature value" field is mapped to the semantic concept of "ambient temperature", and structured data containing semantic tags is generated. Perform context analysis on the preliminary semantic annotation results, extract the association information between adjacent data items, calculate the mutual information value between every two data items, and construct the data item association matrix. The mutual information value is used to measure the correlation between data items. The calculation formula of the mutual information value is:

[0057]

[0058] Among them, I(X;Y) represents the mutual information value between data items X and Y, p(x,y) represents the probability that X and Y take the values ​​of x and y at the same time, and p(x) and p(y) represent the marginal probabilities of data items X and Y respectively. By calculating the mutual information value, the correlation between data items is quantified, and the association matrix of data items is constructed. Based on the association matrix of data items, the vector representation of semantic concepts in the knowledge graph is calculated, and the semantic concept with the highest similarity is selected as the final semantic tag through cosine similarity calculation. According to the final semantic tag, the preliminary semantic annotation results are converted into the Resource Description Framework (RDF) format. RDF is a standardized format for describing resources and their relationships, in which each data item is represented as a subject-predicate-object triple. For example, the data of a temperature sensor is represented as "sensor X" (subject)-"measurement" (predicate)-"temperature value 30 degrees" (object). At the same time, a globally unique URI identifier is assigned to each triple to ensure that these data have unique identity identification in the distributed system, and a data entity with semantic tags is obtained.

[0059] In one example, the sensitivity of standardized data is calculated, and an encryption algorithm is selected according to the sensitivity to encrypt the standardized data to obtain an encrypted security data packet and an encrypted session key, including: extracting features from the standardized data, mapping the data fields into vector representations using a pre-trained word embedding model, and obtaining a data feature vector set; inputting the data feature vector set into a sensitivity assessment model, which is based on a multi-layer perceptron structure and trained using a cross entropy loss function, outputting a sensitivity score for each data field, and obtaining a sensitivity score set; performing cluster analysis on the sensitivity score set, dividing the data fields into different sensitivity levels, and obtaining a data sensitivity classification result, and selecting a corresponding encryption algorithm from a predefined encryption algorithm library according to the data sensitivity classification result. , a lightweight encryption algorithm is used for low-sensitivity data, and a strong encryption algorithm is used for high-sensitivity data, to obtain an encryption algorithm selection scheme; based on the encryption algorithm selection scheme, data fields of different sensitivity levels are encrypted separately, and an encrypted session key is generated at the same time to obtain preliminary encrypted data and an encrypted session key; the preliminary encrypted data is processed in blocks, a unique identifier and version information is added to each data block, an encrypted data packet structure is constructed to obtain structured encrypted data, and integrity protection is performed on the structured encrypted data, an HMAC value is calculated and attached to the data packet, to obtain an integrity-protected encrypted data packet; the encrypted session key is encrypted using an asymmetric encryption algorithm, and the encrypted session key is combined with the integrity-protected encrypted data packet to obtain an encrypted security data packet and an encrypted session key.

[0060] In this example, each field in the standardized data is represented as a semantically related feature vector. Word embedding models (such as Word2Vec, GloVe, etc.) map data fields to vector space through pre-trained models, so that fields with similar meanings or semantics are close in the vector space. For example, fields such as device name, data type, and unit are mapped to vectors through word embedding models to obtain a set of data feature vectors. Through this process, each field in the standardized data is converted to a vector representation, which is represented as:

[0061] V i = {v 1 ,v 2 ,…,v n};

[0062] Among them, V i The vector representation of the i-th data field, v j Represents the value of field i in the jth dimension of the word embedding vector space. The data feature vector set is input into the sensitivity assessment model. The sensitivity assessment model adopts a multi-layer perceptron structure, a feedforward neural network composed of multiple layers of fully connected layers, which is used to process high-dimensional feature vectors. The sensitivity assessment model performs sensitivity analysis on the data field and outputs the sensitivity score of each data field. During the training process, the cross entropy loss function is used to measure the difference between the model output and the true label, thereby guiding the model to update the parameters. The calculation formula of the cross entropy loss function is:

[0063]

[0064] Among them, L represents the cross entropy loss, y i is the true label, is the predicted value of the model, and N represents the number of data samples. By minimizing the cross entropy loss, the model can gradually learn how to distinguish data fields of different sensitivities. After training, the model outputs a sensitivity score for each input data field to form a sensitivity score set. The higher the sensitivity score, the more important the data field is in terms of data security and requires more stringent protection. Cluster analysis is performed on the sensitivity score set to divide the data fields into different sensitivity levels. Cluster analysis uses algorithms such as K-means clustering to classify data fields with similar sensitivity scores into the same category, forming classification results of different levels such as high sensitivity and low sensitivity. For example, temperature data is classified as low sensitivity, while user authentication information is classified as high sensitivity. The clustering results are used to guide the selection of data encryption strategies, thereby obtaining data sensitivity classification results. According to the data sensitivity classification results, the corresponding encryption algorithm is selected from the predefined encryption algorithm library. For low-sensitivity data fields, lightweight encryption algorithms such as RC4 or AES-128 are selected. These algorithms have high computational efficiency and are suitable for protecting data with less stringent security requirements. For highly sensitive data fields, strong encryption algorithms such as AES-256 or RSA are selected to ensure data security and prevent theft or tampering during transmission and storage. The encryption algorithm selection scheme is formulated according to the sensitivity level to effectively strike a balance between security and computational efficiency. Based on the encryption algorithm selection scheme, data fields of different sensitivity levels are encrypted separately, and encryption session keys are generated for each group of encryption to obtain preliminary encrypted data and encryption session keys. The encryption session key is used to decrypt data during data transmission to ensure that only the legitimate recipient can decrypt and use the data. The preliminary encrypted data is processed in blocks, and each data block is attached with a unique identifier and version information to construct an encrypted data packet structure. The purpose of block processing is to improve the flexibility and concurrency of transmission. Each data block can be transmitted independently, thereby improving the utilization of network resources and data transmission efficiency. The identifier is used to mark the order of data blocks in the data packet, and the version information is used to record the version of the data block to facilitate the reorganization and management of the data at the receiving end. In this way, structured encrypted data is obtained. In order to ensure the integrity of the data during transmission, integrity protection is performed on the structured encrypted data. Integrity protection is achieved by calculating the HMAC (key-based message authentication code) value. The calculation formula of HMAC is:

[0065] H=Hash(K⊕opad∥Hash(K⊕ipad∥M));

[0066] Where H represents the calculated HMAC value, K is the encryption key, opad and ipad are the external and internal padding constants respectively, ∥ represents the string concatenation operation, and M is the message content. By appending the HMAC value, the receiver is ensured to verify whether the data has been tampered with during transmission. If the data has been tampered with, the calculated HMAC value will not match, so that the integrity problem of the data can be detected. The encrypted session key is encrypted using an asymmetric encryption algorithm. Asymmetric encryption algorithms such as RSA use public key encryption and private key decryption to ensure that the session key will not be illegally intercepted and decrypted during transmission. By combining the encrypted session key with the integrity-protected encrypted data packet, an encrypted security data packet and an encrypted session key are obtained.

[0067] In one example, a multi-dimensional evaluation is performed on the current network environment of the smart IoT gateway to obtain network evaluation data, and the network evaluation data is input into a deep reinforcement learning model for multi-factor transmission channel selection to obtain an optimal transmission channel combination, including: performing parallel scanning on all network interfaces of the smart IoT gateway, collecting interface status information through the SNMP protocol, and obtaining a network interface list; based on the network interface list, performing a network performance test on each active interface, measuring the round-trip delay using the ICMP protocol, testing the bandwidth through the iPerf tool, measuring the packet loss rate using a UDP traffic generator, and additionally measuring the signal strength and signal-to-noise ratio of the wireless interface to obtain an original network performance data matrix containing multi-dimensional indicators; generating a network interface evaluation vector based on the original network performance data matrix , and according to the network interface evaluation vector, the network interfaces are arranged in descending order, and the top N interfaces are selected as the candidate transmission channel set; the network performance data of the candidate transmission channel set is fused with the historical transmission data in time series to construct a state space matrix, in which each row represents the network state at a time point, and the columns include bandwidth utilization and delay jitter indicators to obtain the current network state representation; the current network state representation is input into the deep reinforcement learning model, which uses a dual DQN structure to calculate the Q value of each possible transmission channel combination through forward propagation to obtain a transmission channel evaluation matrix containing the Q values ​​of all possible combinations; based on the transmission channel evaluation matrix, the probability of each transmission channel combination being selected is calculated, and the combination with the highest probability is selected as the optimal transmission channel combination.

[0068] In this example, the status information of the network interface is collected through SNMP (Simple Network Management Protocol) to obtain the availability and health status of the interface. SNMP is a network management protocol that performs parallel scans on each network interface of the IoT gateway device to identify which interfaces are active and the basic configuration parameters of the interface, such as rate, status (enabled or disabled), etc. With this information, a list of network interfaces of the current IoT gateway is obtained. Based on the obtained list of network interfaces, a network performance test is performed on each active network interface to evaluate their actual performance. The round-trip time (RTT) of each interface is measured using the ICMP protocol. ICMP is a commonly used network tool. The Ping command is used to send a request to the network interface and measure the response time to obtain the round-trip delay. The formula for calculating the round-trip delay is:

[0069] RTT=T response -T request ;

[0070] RTT represents the round-trip time, T response Indicates the time when the response is received, T request Indicates the time to send a request. RTT reflects the delay of the network interface when transmitting data and is an important indicator for evaluating interface performance. Use the iPerf tool to test the bandwidth of each network interface. iPerf is a tool used to measure the maximum TCP or UDP bandwidth. It can generate a continuous data stream and measure the maximum throughput of the network. The bandwidth test reflects the performance of the network interface under high load and helps determine which interface is suitable for the transmission of large amounts of data. The UDP traffic generator is used to measure the packet loss rate. The packet loss rate is calculated by simulating high-frequency UDP data packet sending and counting the number of lost data packets:

[0071]

[0072] Among them, Packet Loss Rate represents the packet loss rate, N lost Indicates the number of lost packets, N sent Represents the total number of data packets sent. Packet loss rate is a key indicator reflecting the stability of the network interface. A low packet loss rate indicates that the network has a high reliability in transmitting data under high load. For wireless interfaces, signal strength and signal-to-noise ratio are measured. Signal strength (RSSI) indicates the received power of the wireless signal, while signal-to-noise ratio (SNR) is the ratio of signal to noise. Through testing, the original network performance data matrix containing multi-dimensional indicators is obtained to describe the multi-faceted performance of each network interface. According to the original network performance data matrix, the performance data is processed to generate an evaluation vector for the network interface. The network interface evaluation vector comprehensively considers the performance indicators of each interface, such as RTT, bandwidth, packet loss rate, RSS and SNR, to indicate the overall performance level of the network interface. The evaluation vector is expressed as:

[0073] E i ={rtt i ,bw i ,plr i ,rssi i ,snr i};

[0074] Among them, E i represents the evaluation vector of the ith interface, rtt i is the round trip delay, bw i is the bandwidth, plr i is the packet loss rate, rssi i is the signal strength, snr i is the signal-to-noise ratio. Based on the evaluation vector, the network interfaces are sorted in descending order, and the top N interfaces are selected as the candidate transmission channel set to ensure that the selected channels have the best transmission performance under the current conditions. After determining the candidate transmission channel set, the historical network performance data is combined for time series fusion to construct the current network status representation. The state space matrix is ​​constructed by performing time series analysis on the real-time performance data and historical performance data of the candidate transmission channel set. Each row of the state space matrix represents the network status at a point in time, and the columns include indicators such as bandwidth utilization, delay, and jitter. The state space matrix can be expressed as:

[0075] S t = {bu t ,rtt t ,j t};

[0076] Among them, S t represents the network status at time point t, bu t is the bandwidth utilization, rtt t is the current round trip delay, j t is network jitter. Through the method of time series fusion, the dynamic change characteristics of network performance over time are captured to describe the current network state. The current network state representation is input into the deep reinforcement learning model to select the optimal transmission channel combination. The deep reinforcement learning model uses a dual DQN structure to reduce the overestimation problem of Q value through two independent Q networks and improve the stability of the model. In the forward propagation process of the model, the current state is evaluated, the Q value of each possible transmission channel combination is calculated, and the transmission channel evaluation matrix containing the Q values ​​of all possible combinations is obtained. The calculation formula of Q value is:

[0077]

[0078] Among them, Q(s t ,a) means in state st The Q value of the selected action a, r t is the current instant reward, γ is the discount factor used to measure the importance of future rewards, max a′ Q(s t+1 ,a′) represents the next state s t+1 The maximum Q value under the current network state. In this way, the performance of each transmission channel combination under the current network state is evaluated, and a transmission channel evaluation matrix containing the Q values ​​of all combinations is output. Based on the transmission channel evaluation matrix, the probability of each transmission channel combination being selected is calculated. By converting the Q value into a selection probability, the priority of each channel combination is quantified. The transmission channel combination with the highest probability is selected as the optimal transmission channel combination, thereby ensuring that data transmission can be carried out in the most efficient and stable manner under the current network conditions.

[0079] Among them, the current network state representation is input into the deep reinforcement learning model. The deep reinforcement learning model uses a dual DQN structure to calculate the Q value of each possible transmission channel combination through forward propagation to obtain a transmission channel evaluation matrix containing the Q values ​​of all possible combinations, including: normalizing the current network state representation, using the Z-score standardization method to convert each dimension of data into a standard normal distribution with a mean of 0 and a standard deviation of 1, and obtaining a standardized network state vector; inputting the standardized network state vector into the main network, which is composed of a multi-layer fully connected neural network, and using the ReLU activation function for nonlinear transformation to obtain the main network output features; inputting the main network output features into the advantage flow network and the value flow network, wherein the advantage flow network outputs the advantage value of each action, and the value flow network outputs the state value, which is realized through two independent fully connected layers to obtain the advantage value vector and the state value scalar; based on the advantage value vector and the state value scalar, using Dueling The Q value calculation formula of DQN, i.e., Q(s,a)=V(s)+(A(s,a)-mean(A(s,a))), calculates the Q value of each possible transmission channel combination to obtain the Q value vector of the main network; inputs the standardized network state vector into the target network, the target network structure is the same as the main network, but the parameter update frequency is lower, and the target network Q value vector is obtained through the same forward propagation process; uses the ε-greedy strategy to select actions, selects the action with the largest Q value with a probability of 1-ε, and randomly selects actions with a probability of ε, where the ε value gradually decays during the training process, and obtains the selected transmission channel combination; combines the main network Q value vector and the target network Q value vector to construct a transmission channel evaluation matrix containing the Q values ​​of all possible transmission channel combinations, where each row in the matrix represents a possible transmission channel combination and each column represents the corresponding Q value; normalizes the transmission channel evaluation matrix, and uses the Softmax function to convert the Q value into a selection probability to obtain the final transmission channel selection probability distribution.

[0080] In one example, based on the optimal transmission channel combination, an encrypted security data packet is sent to the IoT platform, the received data confirmation information is verified based on the consensus algorithm, and a data access status report is output through the data visualization interface and the AI ​​voice assistant, including: according to the optimal transmission channel combination, the encrypted security data packet is fragmented, the large data packet is divided into multiple small data packets, and a sequence number and a checksum are added to each fragment to obtain a data fragmentation set; based on the data fragmentation set, the optimal transmission path of each fragment is calculated, and the fragments are assigned to different transmission channels to obtain a fragmentation transmission plan; according to the fragmentation transmission plan, the data fragments are simultaneously sent to the IoT platform through multi-threaded parallel transmission technology, and the timeout retransmission mechanism is started to record the sending status of each fragment. Obtain the data transmission log; receive the data confirmation information returned by the IoT platform, which contains the serial number and digital signature of the successfully received data fragments, and use the asymmetric encryption algorithm to verify the validity of the digital signature to obtain a preliminary verification result; input the preliminary verification result into the consensus algorithm model based on Byzantine fault tolerance, and the consensus algorithm model based on Byzantine fault tolerance performs multiple rounds of voting in a distributed node network. When more than 2 / 3 of the nodes reach a consensus, the data reception confirmation result is output; based on the data reception confirmation result, the success rate, average delay and throughput of the data transmission are calculated, a data access status report is constructed, and an interactive chart is generated through the data visualization interface. The data access status report is converted into voice output through the AI ​​voice assistant using text-to-speech technology.

[0081] In this example, a large data packet is split into multiple small data packets so that the data can be transmitted in parallel in different transmission channels, improving the efficiency and flexibility of transmission. During the fragmentation process, a sequence number and checksum are attached to each fragment to ensure that the data can be correctly reassembled at the receiving end. The sequence number is used to identify the order of data fragments to ensure that the fragments can be reassembled in the correct order, while the checksum is used to ensure the integrity of the data. The checksum is calculated using simple modulo operations, as shown below:

[0082]

[0083] Among them, C i represents the checksum of the i-th data shard, d ijrepresents the jth byte in data slice i, n is the number of bytes in the slice, and M is a constant used for modulo operation. By calculating the checksum, it is ensured that the data has not been tampered with during transmission. After completing the slice processing, the optimal transmission path is calculated for each slice based on the data slice set. The selection of the transmission path for each slice depends on the real-time status of the network, including multiple indicators such as bandwidth utilization, delay, and packet loss rate. By evaluating these indicators, each slice is assigned to a different transmission channel to obtain a slice transmission plan. Make full use of various transmission resources in the network, reduce the load of a single transmission channel, and improve the overall data transmission efficiency and reliability. According to the slice transmission plan, the data slices are sent to the IoT platform simultaneously through multi-threaded parallel transmission technology. Multi-threaded parallel transmission is an efficient network transmission technology that enables multiple slices to be transmitted simultaneously in different channels, thereby improving the speed and stability of transmission. At the same time, in order to ensure reliable data transmission, the timeout retransmission mechanism is started. The core of the timeout retransmission mechanism is that when a slice confirmation is not received within a certain period of time, the slice will be resent to ensure that all slices can eventually reach the IoT platform successfully. The sending status of each shard, including information such as successful sending and timeout retransmission times, will be recorded to generate a data transmission log. After the data shard is successfully transmitted to the IoT platform, the platform will return data confirmation information. The data confirmation information contains the serial number of the successfully received data shard and a digital signature. The digital signature is used to verify the authenticity and integrity of the data and prevent the data from being maliciously tampered with during transmission. The system uses an asymmetric encryption algorithm (such as the RSA algorithm) to verify the digital signature and obtain a preliminary verification result. The asymmetric encryption algorithm uses the public key to decrypt the digital signature, and then compares it with the calculated hash value after decryption to determine the integrity of the data. The formula for asymmetric encryption is as follows:

[0084] D=E d mod N;

[0085] Among them, D represents the decrypted hash value, E is the digital signature, d is the private key, and N is the modulus shared by the public key and the private key. If the decryption result matches the hash value, the data verification is passed, indicating that the data has not been tampered with. The preliminary verification results are input into the consensus algorithm model based on Byzantine fault tolerance for verification. The Byzantine fault-tolerant consensus algorithm is an algorithm that ensures consensus between nodes in a distributed network. Through multiple rounds of voting, the system can still reach a consensus even if there are some unreliable nodes in the network. In IoT applications, the Byzantine fault-tolerant algorithm performs multiple rounds of voting in a distributed node network. When more than 2 / 3 of the nodes reach a consensus on the data confirmation information, the final data reception confirmation result is output. Key performance indicators such as the success rate of data transmission, average latency, and throughput are calculated based on the confirmation results of data reception. The calculation formula for the success rate of data transmission is as follows:

[0086]

[0087] Among them, Success Rate represents the success rate of data transmission, N received Indicates the number of successfully received data fragments, N sent Indicates the total number of data fragments sent. The success rate reflects the proportion of data that reaches the target platform during the transmission process. Average latency is used to measure the time required for data to be sent and confirmed to be received, and throughput indicates the amount of data successfully transmitted per unit time. These indicators combined can comprehensively evaluate the transmission performance of the network. Based on these key indicators, a data access status report is constructed. Through the data visualization interface, these indicators are generated into interactive charts, allowing users to intuitively view the specific performance of data transmission, such as the trend change of success rate, the distribution of average latency, etc. The data access status report is converted into voice output by the AI ​​voice assistant using text-to-speech (TTS) technology. Users can understand the current data transmission status through voice, so that they can still obtain key information in a timely manner when they cannot view the chart.

[0088] Furthermore, the preliminary verification result is encoded into a 256-bit binary string, a 32-bit checksum is added, and a verification message in a standard format is constructed. It is broadcast to all participating nodes in the distributed node network through the P2P network protocol. After each node receives the broadcast information, it parses the verification message, extracts the preliminary verification result, and generates a 512-bit verification proposal containing a 64-bit node ID, a 64-bit nanosecond timestamp, and a 256-bit preliminary verification result; the message digest of the 512-bit verification proposal is calculated using the SHA-256 algorithm to obtain a 256-bit digest value, and then the digest value is signed using the node's private key through the elliptic curve digital signature algorithm (ECDSA) to generate a 512-bit signature value, which is attached to the verification proposal to form a 1024-bit signed verification proposal; the 1024-bit signed verification proposal is sent to other nodes in the network through the UDP protocol, and a receiving buffer of 10MB is established to receive verification proposals from other nodes, and the public key and signature are extracted from each received verification proposal. , use the ECDSA algorithm to verify the validity of the signature, store the verified proposals in the valid proposal list, and obtain a valid verification proposal set; based on the valid verification proposal set, count the number of different node IDs to obtain the total number of nodes N, calculate the number of valid proposals to obtain the number of online nodes M, and determine whether M is greater than (2N+1) / 3. If it is satisfied, the consensus condition judgment result is true, otherwise it is false; according to the consensus condition judgment result, if it is true, enter the voting stage, each node will perform a bitwise XOR operation on the local verification result and the result in each valid verification proposal received. If the result is 0, it is in favor, otherwise it is against, and a 64-bit voting information containing a 1-bit approval / disapproval flag and 63 bits of random padding data is generated; the 64-bit voting information is input as plaintext into the Paillier homomorphic encryption algorithm, and encrypted using a 2048-bit public key to obtain a 2048-bit ciphertext. This ciphertext is the encrypted voting information, which ensures the privacy of the vote; based on the encrypted voting information, execute the secure multi-party computing protocol, specifically using Shamir's The Secret Sharing scheme divides each encrypted voting information into N shares and distributes them to N nodes in the network. Each node performs homomorphic addition operations on all the shares received locally to obtain a partial sum, and then broadcasts the partial sum to other nodes. Finally, each node collects all the partial sums and merges them to obtain a 4096-bit encrypted voting statistics result. The 4096-bit encrypted voting statistics are decrypted using a (k,n) threshold decryption scheme, where k=2N / 3+1. At least k nodes are required to provide their own private key shares to complete the decryption. Each node participating in the decryption uses its own private key share to partially decrypt the ciphertext, and then broadcasts the partial decryption results. When at least k partial decryption results are collected, the complete decryption result is reconstructed using the Lagrange interpolation method to obtain a 64-bit plaintext voting statistics result.According to the 64-bit plaintext voting statistics, extract the number of votes in favor and determine whether it exceeds 2N / 3+1. If so, generate a 256-bit consensus result including the "consensus reached" status, the current round number and the final result. If not, add 1 to the current round number. If it does not exceed the preset maximum round number of 20, return to step 5 to restart a new round of voting. Otherwise, generate a consensus result of "consensus failed". Based on the final consensus result, construct a data reception confirmation report containing a 256-bit data reception confirmation status, a 64-bit timestamp and a variable length (maximum 1MB) list of participating consensus nodes. Use the Merkle tree structure to hash the report to obtain a 32-byte root hash value. The root hash value is written into the blockchain network as transaction data. At the same time, the complete data reception confirmation report is sent to all participating nodes through a reliable multicast protocol to complete the consensus confirmation process of data reception. ;

[0089] Reference Figure 2 This embodiment provides a data access system for a smart Internet of Things gateway, including:

[0090] Identification unit 1, used to perform feature matching and protocol identification on multiple sensors and terminal devices connected to the smart IoT gateway using a device feature library to obtain a device protocol type;

[0091] Parsing unit 2 is used to activate the corresponding protocol parsing plug-in based on the device protocol type, parse the original data to obtain the parsing result, and call the edge computing microservice to input the parsing result and the original data into the semantic ontology model for data standardization processing to obtain standardized data;

[0092] Encryption unit 3, used to calculate the sensitivity of the standardized data, and select an encryption algorithm according to the sensitivity to encrypt the standardized data to obtain an encrypted security data packet and an encrypted session key;

[0093] A selection unit 4 is used to perform a multi-dimensional evaluation on the current network environment of the smart IoT gateway to obtain network evaluation data, and input the network evaluation data into a deep reinforcement learning model to perform multi-factor transmission channel selection to obtain an optimal transmission channel combination;

[0094] The output unit 5 is used to send the encrypted security data packet to the IoT platform based on the optimal transmission channel combination, verify the received data confirmation information based on the consensus algorithm, and output the data access status report through the data visualization interface and AI voice assistant.

[0095] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0096] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0097] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0098] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0099] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0100] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, system, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, system, article or method. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, system, article or method including the element.

[0101] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A data access method for a smart Internet of Things gateway, characterized in that: The following steps are involved: Use the device feature library to perform feature matching and protocol identification on multiple sensors and terminal devices connected to the smart IoT gateway to obtain the device protocol type; Activate the corresponding protocol parsing plug-in based on the device protocol type, parse the original data to obtain the parsing result, and call the edge computing microservice to input the parsing result and the original data into the semantic ontology model for data standardization processing to obtain standardized data; Calculating the sensitivity of the standardized data, and selecting an encryption algorithm according to the sensitivity to encrypt the standardized data to obtain an encrypted security data packet and an encrypted session key; Perform a multi-dimensional evaluation on the current network environment of the smart IoT gateway to obtain network evaluation data, and input the network evaluation data into a deep reinforcement learning model to perform multi-factor transmission channel selection to obtain an optimal transmission channel combination; Based on the optimal transmission channel combination, the encrypted security data packet is sent to the IoT platform, the received data confirmation information is verified based on the consensus algorithm, and the data access status report is output through the data visualization interface and AI voice assistant.

2. The data access method of the smart Internet of Things gateway according to claim 1 is characterized in that: The device feature library is used to perform feature matching and protocol identification on multiple sensors and terminal devices connected to the smart IoT gateway to obtain the device protocol type, including: Monitor the communication interface of the smart IoT gateway in real time, identify newly connected sensors and terminal devices, and record the physical connection characteristics of the sensors and terminal devices to obtain a list of device connection information; Extracting protocol identifiers, data frame formats and communication parameters of multiple pre-stored communication protocols from the device feature library to construct a protocol feature matrix; Calculate the similarity between the physical connection features in the device connection information list and the protocol feature matrix to obtain a device-protocol similarity score, and select K candidate protocols with the highest similarity for each device based on the device-protocol similarity score to obtain a device candidate protocol set; For each communication protocol in the candidate protocol set of the device, data packet capture and protocol simulation are performed, and the protocol matching degree is calculated by sending simulated data packets to the device and analyzing the response to obtain a protocol matching score table; According to the protocol matching score table, the similarity score and the protocol matching degree are comprehensively considered, and the protocol with the highest votes is selected as the identification result of the device to obtain a preliminary device-protocol mapping relationship; For the devices whose confidence in the preliminary device-protocol mapping relationship is lower than a preset threshold, adaptive protocol learning is performed and key features are extracted, the key features are compared with a known protocol library, and a temporary protocol is generated; The temporary protocol is applied to the real-time data stream of the corresponding device, a protocol parsing test is performed, and the parsing success rate and the data integrity index are calculated. If the parsing success rate and the data integrity index meet the preset conditions, the temporary protocol is added to the device feature library, and the device-protocol mapping relationship is updated to obtain the device protocol type.

3. The data access method of the smart Internet of Things gateway according to claim 2 is characterized in that: The method activates the corresponding protocol parsing plug-in based on the device protocol type, parses the original data to obtain the parsing result, and calls the edge computing microservice to input the parsing result and the original data into the semantic ontology model for data standardization processing to obtain standardized data, including: According to the device protocol type, a matching protocol parsing plug-in is retrieved from a protocol parsing plug-in library, and the protocol parsing plug-in is loaded into a parsing engine of the smart IoT gateway to obtain a protocol parsing plug-in; Inputting the original data into the protocol analysis plug-in, decomposing, extracting and converting the original data using the preset analysis rules to obtain the analysis results, and performing data integrity check on the analysis results, calculating the data checksum, and comparing it with the preset threshold to obtain the data validity mark; The parsing result and the data validity mark are transmitted to the edge computing microservice, and the edge computing task is triggered to obtain the preliminarily processed data; Extracting a semantic ontology model related to the device protocol type from a semantic ontology library, constructing a semantic mapping rule set, and semantically annotating the preliminarily processed data and the original data according to the semantic mapping rule set to generate a data entity with semantic tags; Ontological reasoning is performed on the data entities with semantic tags, and data relationships are analyzed and derived using description logic algorithms to obtain a semantic association network, and normalization processing is performed based on the semantic association network to obtain standardized data.

4. The data access method of the smart Internet of Things gateway according to claim 3 is characterized in that: The extracting of the semantic ontology model related to the device protocol type from the semantic ontology library, constructing a semantic mapping rule set, and semantically annotating the preliminarily processed data and the original data according to the semantic mapping rule set to generate a data entity with semantic tags includes: According to the device protocol type, a query statement is constructed, the query statement includes a protocol identifier and a keyword, a semantic ontology library is searched using a SPARQL query language, an ontology model related to the device protocol type is extracted, and a plurality of candidate ontology models are obtained; Perform structural analysis on each candidate ontology model, extract core concepts, attributes and relationships, construct ontology feature vectors, and calculate feature weights to obtain a weighted ontology feature vector set; Performing similarity matching between the weighted ontology feature vector set and the corresponding domain knowledge graph, and selecting the candidate ontology model with the highest similarity as the most matching semantic ontology model; Based on the semantic ontology model, extracting class, attribute and relationship definitions, and constructing a semantic mapping rule set, the semantic mapping rule set includes a mapping relationship from data fields to semantic concepts, attribute constraints and reasoning rules; Parsing the preliminarily processed data and the original data into key-value pairs, and assigning corresponding semantic concepts and attributes to each key-value pair according to the semantic mapping rule set, to generate a preliminary semantic annotation result; Performing context analysis on the preliminary semantic annotation results, extracting association information between adjacent data items, calculating mutual information values, and constructing a data item association matrix. At the same time, based on the data item association matrix, calculating the vector representation of the semantic concept in the knowledge graph, and selecting the semantic concept with the highest cosine similarity as the final semantic tag; The preliminary semantic annotation result is converted into a resource description framework format according to the final semantic tag, each data item is represented as a subject-predicate-object triple, and a globally unique URI identifier is assigned to each triple to obtain a data entity with a semantic tag.

5. The data access method of the smart Internet of Things gateway according to claim 1 is characterized in that: The calculating the sensitivity of the standardized data and selecting an encryption algorithm according to the sensitivity to encrypt the standardized data to obtain an encrypted security data packet and an encrypted session key includes: Extracting features from the standardized data, mapping the data fields into vector representations using a pre-trained word embedding model, and obtaining a data feature vector set; Inputting the data feature vector set into a sensitivity assessment model, the sensitivity assessment model is based on a multi-layer perceptron structure, trained using a cross entropy loss function, and outputting a sensitivity score for each data field to obtain a sensitivity score set; Performing cluster analysis on the sensitivity score set, dividing the data fields into different sensitivity levels, obtaining a data sensitivity classification result, and selecting a corresponding encryption algorithm from a predefined encryption algorithm library according to the data sensitivity classification result, using a lightweight encryption algorithm for low-sensitivity data and a strong encryption algorithm for high-sensitivity data, to obtain an encryption algorithm selection scheme; Based on the encryption algorithm selection scheme, data fields of different sensitivity levels are encrypted respectively, and encryption session keys are generated at the same time to obtain preliminary encrypted data and encryption session keys; Processing the preliminary encrypted data in blocks, adding a unique identifier and version information to each data block, constructing an encrypted data packet structure to obtain structured encrypted data, performing integrity protection on the structured encrypted data, calculating an HMAC value and appending it to the data packet to obtain an integrity-protected encrypted data packet; The encrypted session key is encrypted using an asymmetric encryption algorithm, and the encrypted session key is combined with the integrity-protected encrypted data packet to obtain an encrypted security data packet and an encrypted session key.

6. The data access method of the smart Internet of Things gateway according to claim 1 is characterized in that: The multi-dimensional evaluation of the current network environment of the smart IoT gateway is performed to obtain network evaluation data, and the network evaluation data is input into a deep reinforcement learning model to perform multi-factor transmission channel selection to obtain an optimal transmission channel combination, including: Perform parallel scanning on all network interfaces of the smart IoT gateway, collect interface status information through the SNMP protocol, and obtain a network interface list; Based on the network interface list, a network performance test is performed on each active interface, the round-trip delay is measured using the ICMP protocol, the bandwidth is tested through the iPerf tool, the packet loss rate is measured using the UDP traffic generator, and the signal strength and signal-to-noise ratio are additionally measured for the wireless interface, so as to obtain an original network performance data matrix containing multi-dimensional indicators; Generate a network interface evaluation vector according to the original network performance data matrix, and arrange the network interfaces in descending order according to the network interface evaluation vector, and select the top N interfaces as a set of candidate transmission channels; The network performance data of the candidate transmission channel set is integrated with the historical transmission data in time series to construct a state space matrix, where each row represents the network state at a time point, and the columns include bandwidth utilization and delay jitter indicators, to obtain a representation of the current network state; Inputting the current network state representation into a deep reinforcement learning model, the deep reinforcement learning model uses a dual DQN structure to calculate the Q value of each possible transmission channel combination through forward propagation to obtain a transmission channel evaluation matrix containing all possible combination Q values; Based on the transmission channel evaluation matrix, the probability of each transmission channel combination being selected is calculated, and the combination with the highest probability is selected as the optimal transmission channel combination.

7. The data access method of the smart Internet of Things gateway according to claim 1 is characterized in that: Based on the optimal transmission channel combination, the encrypted security data packet is sent to the IoT platform, the received data confirmation information is verified based on the consensus algorithm, and a data access status report is output through the data visualization interface and the AI ​​voice assistant, including: According to the optimal transmission channel combination, the encrypted secure data packet is fragmented, a large data packet is divided into multiple small data packets, and a sequence number and a checksum are added to each fragment to obtain a data fragment set; Based on the data shard set, calculate the optimal transmission path for each shard, allocate the shards to different transmission channels, and obtain a shard transmission plan; According to the shard transmission scheme, the data shards are sent to the IoT platform simultaneously through multi-threaded parallel transmission technology, and the timeout retransmission mechanism is started to record the sending status of each shard and obtain a data transmission log; Receive the data confirmation information returned by the IoT platform, which contains the successfully received data segment serial number and digital signature, and use the asymmetric encryption algorithm to verify the validity of the digital signature to obtain a preliminary verification result; Input the preliminary verification result into a consensus algorithm model based on Byzantine fault tolerance, and the consensus algorithm model based on Byzantine fault tolerance performs multiple rounds of voting in a distributed node network. When more than 2 / 3 of the nodes reach a consensus, the data reception confirmation result is output; Based on the data reception confirmation result, the success rate, average delay and throughput of data transmission are calculated, a data access status report is constructed, and an interactive chart is generated through a data visualization interface. The data access status report is converted into voice output through an AI voice assistant using text-to-speech technology.

8. A data access system for a smart Internet of Things gateway, characterized in that: For implementing the steps of the method according to any one of claims 1 to 7, the system comprises: An identification unit is used to perform feature matching and protocol identification on multiple sensors and terminal devices connected to the smart IoT gateway using a device feature library to obtain a device protocol type; A parsing unit, configured to activate a corresponding protocol parsing plug-in based on the device protocol type, parse the original data to obtain a parsing result, and call an edge computing microservice to input the parsing result and the original data into a semantic ontology model for data standardization processing to obtain standardized data; An encryption unit, used to calculate the sensitivity of the standardized data, and select an encryption algorithm according to the sensitivity to encrypt the standardized data to obtain an encrypted security data packet and an encrypted session key; A selection unit, configured to perform a multi-dimensional evaluation on the current network environment of the smart IoT gateway to obtain network evaluation data, and input the network evaluation data into a deep reinforcement learning model to perform multi-factor transmission channel selection to obtain an optimal transmission channel combination; The output unit is used to send the encrypted security data packet to the IoT platform based on the optimal transmission channel combination, verify the received data confirmation information based on the consensus algorithm, and output the data access status report through the data visualization interface and AI voice assistant.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images