Access method, device, storage medium and electronic equipment of platform

Abstract

The application discloses an access method and device of a platform, a storage medium and electronic equipment, and relates to the technical field of platforms. The method comprises the following steps: determining a target protocol field included in original data; determining target original data associated with an initial semantic label from the original data based on each initial semantic label, determining target physical state data associated with the initial semantic label from physical state data, and determining target operation logs associated with the initial semantic label from operation logs; determining a target weight of the target protocol field under each initial semantic label based on the target original data, the target physical state data and the target operation logs, and determining a target semantic label of the target protocol field from the multiple initial semantic labels based on the target weight; and accessing a target device to a target platform based on the target semantic label. The method solves the problem that non-standard Internet of Things devices cannot access an Internet of Things platform, and achieves the effect of enabling non-standard Internet of Things devices to access an Internet of Things platform.

Description

Technical Field

[0001] This application relates to the field of computers, and more specifically, to a platform access method, apparatus, storage medium, and electronic device. Background Technology

[0002] In related technologies, due to the different usage environments and manufacturers of various IoT devices on the market, the protocols used when building IoT platforms cannot be unified. In addition, some IoT devices lack interface documentation and development toolkits, and some IoT devices do not adopt well-known or industry-standard communication protocols. Therefore, unified access cannot be achieved.

[0003] This indicates that there is a technical problem in the relevant technologies that prevents non-standard IoT devices from accessing the IoT platform.

[0004] The target has not yet proposed an effective solution to the aforementioned problems in the relevant technologies. Summary of the Invention

[0005] This application provides a platform access method, apparatus, storage medium, and electronic device to at least solve the technical problem in the related art that non-standard IoT devices cannot access the IoT platform.

[0006] According to one aspect of the embodiments of this application, a platform access method is provided, comprising: acquiring raw data, physical status data, and operation logs transmitted by a target device; determining a target protocol field included in the raw data; matching a plurality of initial semantic tags for the target protocol field; determining target raw data associated with each initial semantic tag from the raw data, determining target physical status data associated with each initial semantic tag from the physical status data, and determining target operation logs associated with each initial semantic tag from the operation logs; determining a target weight of the target protocol field under each initial semantic tag based on the target raw data, the target physical status data, and the target operation logs, and determining a target semantic tag of the target protocol field from the plurality of initial semantic tags based on the target weight; and accessing the target device to a target platform based on the target semantic tag.

[0007] In an exemplary embodiment, determining the target weight of the target protocol field under each initial semantic tag based on the target raw data, target physical state data, and the target operation log includes: performing the following operations for any initial semantic tag to obtain the target weight of the initial semantic tag: under the initial semantic tag, determining a first variance of the target protocol field based on the target raw data, determining a second variance of the target protocol field based on the target physical state data, and determining a third variance of the target protocol field based on the target operation log; and determining the target weight based on the first variance, the second variance, and the third variance.

[0008] In an exemplary embodiment, determining the target weight based on the first variance, the second variance, and the third variance includes: determining a first reciprocal of the first variance, a second reciprocal of the second variance, and a third reciprocal of the third variance; determining a first sum of the first reciprocal, the second reciprocal, and the third reciprocal; determining a first ratio of the first reciprocal to the first sum as a first target sub-weight, a second ratio of the second reciprocal to the first sum as a second target sub-weight, and a third ratio of the third reciprocal to the first sum as a third target sub-weight; and determining the first target sub-weight, the second target sub-weight, and the third target sub-weight as the target weight.

[0009] In one exemplary embodiment, determining the first variance of the target protocol field based on the target raw data includes: splitting the target raw data into multiple target data packets; determining the target number of the target data packets; determining the value of the target protocol field included in each target data packet to obtain multiple first values; determining the arithmetic mean of the multiple first values; and determining the first variance based on the arithmetic mean and the first values.

[0010] In an exemplary embodiment, determining the first variance based on the arithmetic mean and the first numerical value includes: determining a first difference between each of the first numerical values ​​and the arithmetic mean; determining the square of each of the first differences to obtain a first squared value; determining the sum of the first squared values ​​to obtain a second sum; determining a fourth reciprocal of the second difference between the target quantity and a first constant; determining the arithmetic square root of the first product of the fourth reciprocal and the second sum as a first index; determining a second product of the first index, a first adaptive parameter, and a second constant; and determining the second power of the natural constant as the first variance.

[0011] In an exemplary embodiment, determining the second variance of the target protocol field based on the target physical state data includes: determining an observation period for observing the target physical state data, and dividing the observation period into multiple sub-observation period periods; performing the following operations for each sub-observation period period to determine the second square value corresponding to the sub-observation period period: determining a first change in the target physical state data under the sub-observation period period and a second change in the second value of the target protocol field under the sub-observation period period; determining a first norm of the first change and a second norm of the second change; determining a third product of a second adaptive parameter and the second norm; determining a third difference between the first norm and the third product; determining the square of the third difference to obtain the second square value; determining the sum of the second square values ​​to obtain a third sum; determining a fourth ratio of the third sum to the total number of sub-observation period periods; determining the arithmetic square root of the fourth ratio as a second index; determining a fourth sum of the second index and a first constant; and determining a fifth ratio of the second index to the fourth sum as the second variance.

[0012] In one exemplary embodiment, determining the third difference of the target protocol field based on the target operation log includes: determining a first number of times the target device receives a target operation instruction based on the target operation log; determining a second number of times the target device responds to the target operation instruction; and determining the third difference based on the first number and the second number.

[0013] In an exemplary embodiment, determining the third difference based on the first number and the second number includes: determining a sixth ratio of the second number to the first number; determining a fourth difference between the indicator function of each target operation instruction and the sixth ratio; determining the square of each fourth difference to obtain a third square value; determining a fifth sum of each third square value; determining the arithmetic square root of the seventh ratio of the first constant to the fifth sum as a third index; and determining the hyperbolic tangent function value of the fourth product of the third index and the third adaptive parameter as the third difference.

[0014] In an exemplary embodiment, determining the target semantic label of the target protocol field from a plurality of initial semantic labels based on the target weight includes: performing the following operations for each initial semantic label to determine a comparison result corresponding to the initial semantic label: determining a target filtering strategy based on the target weight; determining target data based on the target filtering strategy, wherein the target data includes the target original data, the target physical state data, and data included in the target operation log; determining a third value corresponding to the target data; and determining a fourth value of the initial semantic label stored in the target database; comparing the third value with the fourth value to obtain the comparison result; determining candidate semantic labels based on the comparison result; and determining the target semantic label based on the candidate semantic labels.

[0015] In an exemplary embodiment, determining a candidate semantic label based on the comparison result includes: determining a target comparison result included in the comparison result, wherein the target comparison result indicates that the third value is the same as the fourth value; and determining the initial semantic label corresponding to the target comparison result as the candidate semantic label.

[0016] In one exemplary embodiment, determining a target filtering strategy based on the target weight includes: if a second target sub-weight included in the target weight is greater than a first preset threshold, determining the target filtering strategy as a device behavior-driven strategy, wherein the second target sub-weight is the weight of the target physical state data; if the second target sub-weight included in the target weight is less than the first preset threshold and a third target sub-weight is greater than the second preset threshold, determining the target filtering strategy as an operation instruction-driven strategy, wherein the third target sub-weight is the weight of the target operation log; if the second target sub-weight included in the target weight is less than the first preset threshold and the third target sub-weight is less than the second preset threshold, determining the target filtering strategy as a network traffic-driven strategy.

[0017] In one exemplary embodiment, determining target data based on the target filtering strategy includes: determining the target data as target physical state data when the target filtering strategy is a device behavior-driven strategy; determining the target data as target operation logs when the target filtering strategy is an operation instruction-driven strategy; and determining the target data as target raw data when the target filtering strategy is a network traffic-driven strategy.

[0018] In an exemplary embodiment, determining the target semantic label based on the candidate semantic labels includes: performing the following operations for each candidate semantic label to determine a target confidence level for the candidate semantic label; determining the target semantic label based on the target confidence level: determining candidate raw data, candidate physical state data, and candidate operation logs corresponding to each candidate semantic label; determining a first objective function based on the candidate raw data, the first objective function representing the strength of evidence that the target protocol field is labeled as the candidate semantic label in a first modality; determining a second objective function based on the candidate physical state data, the second objective function representing the strength of evidence that the target protocol field is labeled as the candidate semantic label in a second modality; determining a third objective function based on the candidate operation logs, the third objective function representing the strength of evidence that the target protocol field is labeled as the candidate semantic label in a third modality; and determining the target confidence level based on the first objective function, the second objective function, and the third objective function.

[0019] In an exemplary embodiment, determining the target confidence level based on the first objective function, the second objective function, and the third objective function includes: determining a fifth product of a first target sub-weight included in the target weights and the first objective function; determining a sixth product of a second target sub-weight included in the target weights and the second objective function; determining a seventh product of a third target sub-weight included in the target weights and the third objective function; and determining the sixth sum of the fifth product, the sixth product, and the seventh product as the target confidence level.

[0020] In an exemplary embodiment, determining a first objective function based on the candidate raw data and the candidate semantic labels includes: determining a first feature of the candidate raw data under the candidate semantic labels and a second feature of the candidate semantic labels in a target database; determining a fifth difference between the first feature and the second feature; determining the square of the fifth difference to obtain a fourth squared value; determining the square of a predetermined deviation coefficient to obtain a fifth squared value; determining an eighth product of a third constant and the fifth squared value; determining an eighth ratio of the fifth difference to the eighth product; and determining the exponential function of the eighth ratio as the first objective function.

[0021] In one exemplary embodiment, based on the candidate physical state data and determining a second objective function, the method includes: determining a first change sequence of the target protocol field at adjacent time points; determining a second change sequence of the candidate physical state data at adjacent time points; and determining the absolute value of the Pierre correlation coefficient of the first change sequence and the second change sequence as the second objective function.

[0022] In one exemplary embodiment, determining a third objective function based on the candidate operation log includes: determining a first number of times the target device receives a target operation instruction; determining a second number of times the target device responds to the target operation instruction; and determining a ninth ratio of the first number and the second number as the third objective function.

[0023] In an exemplary embodiment, determining the target semantic label based on the target confidence level includes: determining the candidate semantic label corresponding to the maximum confidence level included in the target confidence level as the target semantic label; or, determining a sixth difference between each pair of target confidence levels, and if the sixth difference includes a seventh difference less than a third preset threshold, determining the target weight corresponding to the seventh difference, and determining the candidate semantic label corresponding to the maximum target weight as the target semantic label; or, if the sixth difference includes a seventh difference less than a third preset threshold, determining the similarity between the candidate semantic labels corresponding to the target confidence level of the seventh difference, merging the candidate semantic labels with similarity greater than a fourth preset threshold to obtain a merged semantic label, and determining the merged semantic label as the target semantic label.

[0024] In an exemplary embodiment, before acquiring the raw data, physical status data, and operation log transmitted by the target device, the method further includes: acquiring the raw byte stream of the target device; removing the header information and metadata of the data packets included in the raw byte stream to obtain a standardized byte stream; and injecting target metadata into the standardized byte stream to obtain the raw data, wherein the target metadata is known protocol metadata.

[0025] According to another aspect of the embodiments of this application, a platform access device is also provided, comprising: an acquisition module, configured to acquire raw data transmitted by a target device, physical status data, and operation logs; a first determination module, configured to determine a target protocol field included in the raw data; a matching module, configured to match a plurality of initial semantic tags for the target protocol field; a second determination module, configured to determine target raw data associated with each initial semantic tag from the raw data, target physical status data associated with the initial semantic tag from the physical status data, and target operation logs associated with the initial semantic tags from the operation logs; a third determination module, configured to determine a target weight of the target protocol field under each initial semantic tag based on the target raw data, the target physical status data, and the target operation logs, and to determine a target semantic tag of the target protocol field from a plurality of initial semantic tags based on the target weight; and an access module, configured to access the target device to the target platform based on the target semantic tag.

[0026] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed by a processor.

[0027] According to another aspect of the embodiments of this application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform the steps in any of the method embodiments described above.

[0028] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to perform the steps of any of the above method embodiments through the computer program.

[0029] This application first obtains the raw data, physical status data, and operation logs transmitted by the target device. Multiple target protocol fields can be identified from the raw data, and multiple initial semantic tags can be assigned to each target protocol field. Based on different initial semantic tags, target raw data associated with a specific initial semantic tag can be identified from the raw data, target physical status data associated with a specific initial semantic tag can be identified from the physical status data, and target operation logs associated with a specific initial semantic tag can be identified from the operation logs. Different target weights for target protocol fields under different initial semantic tags can be determined using the target raw data, target physical status data, and target operation logs. Using these target weights, unique target semantic tags for target protocol fields can be filtered from the initial semantic tags. Subsequently, the target device can be connected to the target platform using these target semantic tags. Since semantic analysis can be performed from multiple modalities (raw data, physical status data, and operation logs), even without protocol documentation, semantic tags can be identified by comparing target protocol fields with device behavior (physical status data) and the effects of user operations (operation command logs). Therefore, the problem of non-standard IoT devices being unable to connect to IoT platforms can be solved, achieving the goal of enabling non-standard IoT devices to access IoT platforms. Attached Figure Description

[0030] Figure 1 This is a schematic diagram illustrating an application scenario of a platform access method according to an embodiment of this application;

[0031] Figure 2 This is a flowchart illustrating an optional platform access method according to an embodiment of this application;

[0032] Figure 3 This is a system architecture diagram of an IoT device protocol parsing system based on multimodal feature fusion according to an embodiment of this application;

[0033] Figure 4 This is a schematic diagram illustrating the driver generation according to an embodiment of this application;

[0034] Figure 5 This is a schematic diagram illustrating closed-loop calibration according to an embodiment of this application;

[0035] Figure 6 This is a schematic diagram illustrating the closed-loop protocol correction layer according to an embodiment of this application;

[0036] Figure 7 It is an automatic signal path according to the embodiments of this application;

[0037] Figure 8 This is a manually triggered path according to the embodiments of this application;

[0038] Figure 9These are the fusion variance calculation steps according to the embodiments of this application;

[0039] Figure 10 This is a flowchart of the protocol dynamic field filtering process according to an embodiment of this application;

[0040] Figure 11 This is a flowchart illustrating a platform access method according to a specific embodiment of this application;

[0041] Figure 12 This is a structural block diagram of an optional platform access device according to an embodiment of this application. Detailed Implementation

[0042] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0043] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0044] According to one aspect of the embodiments of this application, a platform access method is provided. Optionally, in this embodiment, the above-described platform access method may be applied to, but is not limited to, platforms such as... Figure 1 The hardware environment shown includes terminal device 102 and server 104. Server 104 can be connected to terminal device 102 via a network and can be used to provide services (e.g., application services, etc.) to terminal device 102 or clients installed on terminal device 102. A database can be set up on server 104 or independently of server 104 to provide data storage services for server 104.

[0045] The aforementioned network may include, but is not limited to, at least one of the following: wired network and wireless network. The aforementioned wired network may include, but is not limited to, at least one of the following: wide area network (WAN), metropolitan area network (MAN), and local area network (LAN). The aforementioned wireless network may include, but is not limited to, at least one of the following: Wireless Fidelity (WIFI) and Bluetooth. Terminal device 102 may be, but is not limited to, a personal computer (PC), mobile phone, tablet computer, etc. Server 104 may be, but is not limited to, a cloud server, server cluster, or other server types.

[0046] The platform access method of this application embodiment can be executed by server 104, terminal device 102, or jointly by server 104 and terminal device 102. Alternatively, the terminal device 102 can execute the platform access method of this application embodiment by a client installed on it.

[0047] Figure 2 This is a flowchart illustrating an optional platform access method according to an embodiment of this application, such as... Figure 2 As shown, the process of this method may include the following steps:

[0048] Step S202: Obtain the raw data, physical status data, and operation logs transmitted by the target device;

[0049] The platform access method in this embodiment can be applied to the Internet of Things (IoT) field, specifically to device access scenarios during IoT construction. IoT devices typically need to adhere to certain communication protocols when connecting to an IoT platform to ensure accurate data transmission and interoperability between devices. However, different IoT devices use different protocols. When parsing the communication protocols of non-standard devices, it is often difficult to accurately annotate the semantics of fields. Furthermore, in the absence of clear protocol documentation, the meaning of fields is often misinterpreted or misunderstood, thus hindering the access of IoT devices.

[0050] To at least partially solve the above-mentioned technical problems, in this embodiment, the specific meaning of the protocol fields can be analyzed from a multimodal perspective: Figure 3 This is a system architecture diagram of an IoT device protocol parsing system based on multimodal feature fusion according to an embodiment of this application, such as... Figure 3As shown, the data acquisition layer can obtain the raw byte stream of the target device (IoT device) through the physical layer interface. This can be achieved through packet capture using common packet capture tools such as tcpdump and Wireshark, enabling automated encapsulation and real-time copying of the byte stream. The target device can be an IoT device, which can be understood as a physical device capable of connecting to the internet and collecting, sending, or receiving data through sensors, actuators, or other technologies. This includes environmental sensors, smart home appliances, wearable devices, medical devices, etc., but is not limited to these. The protocol processing layer can automatically identify the raw byte stream using a protocol unpacking engine, achieving unified payload extraction and protocol feature injection. This results in a standardized byte stream that generates network traffic state data (i.e., the raw data mentioned above). The raw byte stream can be understood as the most basic form of data transmitted over a network or between devices without parsing or processing. It contains all the information sent and received by the target device, including various protocol headers, payload data, checksums, etc. Data is often transmitted in binary form, while the protocol defines how to interpret and process these byte streams. The data acquisition layer can also collect physical state data of the target device through device behavior collectors (such as additional IoT or visual sensing devices) to generate device behavior data (i.e., the physical state data mentioned above), which is consistent with the data reported by the target device, similar to the mirror principle; the data acquisition layer can also obtain the user's operation actions and operation process on the control terminal through the user operation collector, and convert them into operation instruction logs to generate user operation data (i.e., the operation log mentioned above).

[0051] Step S204: Determine the target protocol field included in the original data;

[0052] In the above embodiments, since the IoT protocol is composed of multiple parts, each of which carries specific information, the fields in a data packet can be arranged and encoded according to the protocol to ensure that the data can be correctly transmitted and interpreted. Therefore, multiple target protocol fields can be directly obtained from the raw data. For example, an MQTT (Message Queuing Telemetry Transport) protocol data packet may contain the following fields: Fixed header: contains basic information such as control packet type and remaining length; Topic: used to specify the destination of the message; Message ID: used in messages that require a response, such as publish confirmation or subscribe response; Payload: the part that carries the specific application data. A CoAP (Constrained Application Protocol) protocol data packet may include: Header: includes fields such as version, type, token, options, and message ID; Payload: stores the actual data of the CoAP request or response; Options: a variable-length field used to carry additional information such as URI (Uniform Resource Identifier).

[0053] Step S206: Match multiple initial semantic tags for the target protocol field;

[0054] In the above embodiments, after identifying multiple target protocol fields from the raw data, features of these target protocol fields can be extracted, such as location, size, frequency of occurrence, and change patterns in data packets. For each target protocol field, its characteristics can be compared with a typical feature library of known semantic tags (such as temperature, humidity, on / off status, error codes, etc.). Based on the degree of matching, possible semantic tags (i.e., the initial semantic tags mentioned above) are assigned to each target protocol field. Since this stage is the initial confirmation stage, there will be uncertainty and multiple matches; therefore, one target protocol field may correspond to multiple initial tags.

[0055] Step S208: Based on each initial semantic tag, determine the target original data associated with the initial semantic tag from the original data, determine the target physical state data associated with the initial semantic tag from the physical state data, and determine the target operation log associated with the initial semantic tag from the operation log;

[0056] In the above embodiments, for the initial semantic tags of the initially determined target protocol fields, the target raw data matching each initial semantic tag in the data packet can be identified from the raw data, and the control instructions matching each initial semantic tag (i.e., the target operation log mentioned above) can be determined from the operation instructions. For example, if the "switch status" field changes after the user sends the "turn on the light" command, it can indicate that there is a causal relationship between the two. It can also be determined from the physical state data whether the changes of the target protocol fields of each initial semantic tag are synchronized. The data that changes synchronously can be identified as the target physical state data. For example, if the temperature sensor reading changes accordingly when the "temperature" field changes, it can indicate that there is a correlation between the two.

[0057] Step S210: Based on the target original data, the target physical state data, and the target operation log, determine the target weight of the target protocol field under each initial semantic tag, and determine the target semantic tag of the target protocol field from multiple initial semantic tags based on the target weight;

[0058] In the above embodiments, for each initial semantic tag of the target protocol field, the target weight can be calculated using the target original data, target physical state data, and target operation log. That is, multiple target weights corresponding to multiple initial semantic tags can be obtained. The target weights can include the weights determined using the target original data, the weights determined using the target physical state data, and the weights determined using the target operation log. The semantic tag corresponding to the weight with the largest fused target weight can be selected from the multiple target weights as the final semantic tag of this target protocol field (i.e., the target semantic tag).

[0059] Step S212: Connect the target device to the target platform based on the target semantic tag.

[0060] In the above embodiments, after obtaining the target semantic tags of the target protocol fields based on multimodal fusion analysis, a standard protocol description file (DSL, Domain-Specific Language) can be generated based on the target semantic tags. This DSL can include the structure of the device data packets, the position of the fields, and the semantic meaning of the fields. For example, when network traffic data captures `0xA001` (on) and `0xA000` (off); device behavior data shows a current sensor detecting a change in current from 0.1A to 1A; user operation data shows the APP clicking the "on" button; and the target semantic tags after fusion parsing show that the occurrence time of current and `0xA001` is aligned (within 100ms), it is labeled as an "on command"; and when no current is received, `0xA000` is labeled as a "off command", then the generated protocol description is {"device":"light","packet":{"b\0xA0\x01":{"state":"ON"},"b\0xA0\x00":{"state":"OFF"}}}. Figure 4 This is a schematic diagram illustrating the driver generation according to an embodiment of this application, such as... Figure 4 As shown, after generating the protocol description file, the protocol description parsing engine can parse the file and generate driver code that can be loaded and executed by the IoT platform (i.e., the target platform mentioned above). It can also define a set of annotated protocol driver interfaces, supporting implementations in multiple target languages. The engine automatically assembles and generates implementations in the target languages ​​using a factory pattern, and then packages them into a program package that can be loaded and run by the target platform. The platform access layer can load driver instances using a microkernel plug-in approach, running them as independent sub-service routines while waiting for device connections.

[0061] In the above embodiments, a drive monitoring feedback closed-loop mechanism can also be established. Figure 5 This is a schematic diagram illustrating closed-loop calibration according to an embodiment of this application. Figure 6 This is a schematic diagram illustrating the closed-loop protocol correction layer according to an embodiment of this application, as shown below. Figure 5 Figure 6 As shown, the protocol correction layer can be configured to configure a driver monitoring system to monitor the data reported by the target device and the parsing results in real time. Based on the matching situation, it can statistically analyze the parsing success rate and field coverage. Furthermore, it can incorporate manual review feedback, feeding back automatic correction signals and manual correction instructions from the automatic correction engine and manual feedback interface to the trimodal semantic analysis layer for dynamic weight adjustment, thereby achieving a closed-loop automatic correction of the device protocol. The workflow of the automatic correction engine can be found in [link to documentation]. Figure 7 , Figure 7 It is an automatic signal path according to the embodiments of this application, such as Figure 7As shown, signal adjustment can be triggered when the parsing error continuously exceeds a user-defined threshold. The workflow of the manual feedback interface can be found in [link to documentation]. Figure 8 , Figure 8 This is a manually triggered path according to the embodiments of this application, such as... Figure 8 As shown, signal adjustments can be triggered after manual review of monitoring indicators and automatic verification records confirms mislabeling. Through closed-loop correction design, this system possesses autonomous evolution capabilities, ultimately achieving industrial-grade accuracy of >99% protocol fidelity, far exceeding traditional static analysis schemes.

[0062] This application first obtains the raw data, physical status data, and operation logs transmitted by the target device. Multiple target protocol fields can be identified from the raw data, and multiple initial semantic tags can be assigned to each target protocol field. Based on different initial semantic tags, target raw data associated with a specific initial semantic tag can be identified from the raw data, target physical status data associated with a specific initial semantic tag can be identified from the physical status data, and target operation logs associated with a specific initial semantic tag can be identified from the operation logs. Different target weights for target protocol fields under different initial semantic tags can be determined using the target raw data, target physical status data, and target operation logs. Using these target weights, unique target semantic tags for target protocol fields can be filtered from the initial semantic tags. Subsequently, the target device can be connected to the target platform using these target semantic tags. Since semantic analysis can be performed from multiple modalities (raw data, physical status data, and operation logs), even without protocol documentation, semantic tags can be identified by comparing target protocol fields with device behavior (physical status data) and the effects of user operations (operation command logs). Therefore, the problem of non-standard IoT devices being unable to connect to IoT platforms can be solved, achieving the goal of enabling non-standard IoT devices to access IoT platforms.

[0063] Optionally, the entity performing the above steps may be a background processor, or other devices with similar processing capabilities, or a machine that integrates at least a data processing device. The data processing device may include, but is not limited to, terminals such as computers and mobile phones.

[0064] In an exemplary embodiment, determining the target weight of the target protocol field under each initial semantic tag based on the target raw data, target physical state data, and the target operation log includes: performing the following operations for any initial semantic tag to obtain the target weight of the initial semantic tag: under the initial semantic tag, determining a first variance of the target protocol field based on the target raw data, determining a second variance of the target protocol field based on the target physical state data, and determining a third variance of the target protocol field based on the target operation log; and determining the target weight based on the first variance, the second variance, and the third variance.

[0065] In the above embodiments, the mapping from binary streams to physical meaning is the core issue in solving the protocol parsing engineering of IoT devices. That is, the three-modal fusion analysis for solving this problem is indispensable. Table 1 explains the necessity of missing three modalities. As shown in Table 1, if the network traffic modality is missing, the original data foundation of the protocol will be lost; if the device behavior modality is missing, it will be impossible to distinguish whether `0x01` represents a temperature value or an error code; if the operation command modality is missing, it will be impossible to construct the causal chain between control commands and traffic response packets.

[0066] In the above embodiments, the theoretical feasibility of trimodal fusion analysis can be understood as the trimodal mutual information content covering the protocol semantic information entropy, i.e., I (protocol semantics) ≤ I (network traffic) + I (device behavior) + I (user operation). The engineering feasibility of trimodal fusion analysis can be seen in Table 2. As shown in Table 2, network traffic data can be obtained using MQTT sensors via serial port monitoring or TCP (Transmission Control Protocol) packet capture; device behavior data can also be obtained using external sensors such as smart socket current detection, smart light lamp detection, and network camera IPC (IP Camera); application operation logs can be obtained through access control management platforms via APP operation logs and system logs; application operation logs can also be obtained through IoT (Internet of Things Devices) device management platforms; and user operation data can be obtained through screen recording analysis. It should be noted that the above methods are merely illustrative examples, and this invention does not limit them.

[0067] Table 1

[0068] Modal missing Serious consequences No network traffic mode Loss of the original data foundation of the protocol Deviceless Behavior Mode Unable to distinguish whether `0x01` represents a temperature value or an error code. No operation instruction mode Unable to construct a causal chain between control commands and flow response packets.

[0069] Table 2

[0070] Modal Data availability Example device Network traffic Serial port monitoring / TCP packet capture MQTT sensor Device behavior External sensor Smart socket current detection, smart light lamp light detection User Operations APP operation logs and screen recording analysis Access control management platform

[0071] In the above embodiments, each initial semantic tag matched in the target protocol field can be used to calculate a target weight; that is, one initial semantic tag corresponds to one target weight. First, the network traffic state variance can be determined using the original target data. (i.e., the first variance mentioned above), the variance of the device behavior is determined through the target physical state data. (i.e., the second variance mentioned above) and the user operation state variance determined through the target operation log. (i.e., the aforementioned third variance). By fusing these three variances, the target weight for each row can be reviewed. For example, the weighted sum of the three variances can be calculated, and the weight of the weighted sum can be determined as the target weight. Through this embodiment, by combining information from different data sources, the system can more comprehensively and accurately understand the meaning of protocol fields and dynamically adjust the weights of these modalities in the parsing process based on the variance of each modality of data. This allows the system to automatically adjust its parsing strategy when faced with changes in data quality or availability, thereby improving the robustness of parsing.

[0072] In an exemplary embodiment, determining the target weight based on the first variance, the second variance, and the third variance includes: determining a first reciprocal of the first variance, a second reciprocal of the second variance, and a third reciprocal of the third variance; determining a first sum of the first reciprocal, the second reciprocal, and the third reciprocal; determining a first ratio of the first reciprocal to the first sum as a first target sub-weight, a second ratio of the second reciprocal to the first sum as a second target sub-weight, and a third ratio of the third reciprocal to the first sum as a third target sub-weight; and determining the first target sub-weight, the second target sub-weight, and the third target sub-weight as the target weight.

[0073] In the above embodiments, the formula for determining the target weight using the first variance, the second variance, and the third variance is as follows: Where m can include p: network traffic state, s: device behavior state, and c: user operation state; This can be understood as the reciprocal of the three-mode variance, that is (i.e., the first reciprocal of the first variance mentioned above) (i.e., the second reciprocal of the second variance mentioned above) and (That is, the third reciprocal of the aforementioned third difference). It can be understood as as well as The sum of these is the first sum mentioned above. Therefore, the first target sub-weight included in the target weight is... (That is, the first ratio of the first reciprocal to the first sum); the second objective sub-weight is... (That is, the second ratio of the second reciprocal to the first sum); the third objective sub-weight is... (That is, the third ratio of the third reciprocal to the first sum). Through this embodiment, by calculating the reciprocal of the variance of three modalities of data—network traffic, device behavior, and user operation—the contribution of each modality in determining the semantics of a field can be quantified.

[0074] In one exemplary embodiment, determining the first variance of the target protocol field based on the target raw data includes: splitting the target raw data into multiple target data packets; determining the target number of the target data packets; determining the value of the target protocol field included in each target data packet to obtain multiple first values; determining the arithmetic mean of the multiple first values; and determining the first variance based on the arithmetic mean and the first values.

[0075] In the above embodiments, the calculation process for the fusion variance weight can be found in [reference needed]. Figure 9 , Figure 9 The fusion variance calculation steps are based on the embodiments of this application, such as... Figure 9 As shown, the basic indicators (field dispersion index D) can be calculated first. p Physical synchronization residual R s and causal deviation C d The fusion variance (the first, second, and third variances after fusion) is calculated using basic indicators, and then the final weight (i.e., the target weight mentioned above) is determined using the fusion variance.

[0076] In the above embodiment, the first variance can be expressed by the field dispersion index D. p Calculate the field dispersion index D p The variance can be calculated using a first value and an arithmetic mean. The field dispersion index can be understood as the degree of dispersion of a protocol field within consecutive data packets based on network traffic, reflecting the stability of the protocol field (the smaller the dispersion, the more stable the field). First, the target data can be split into multiple target data packets, and the value of the target protocol field included in each data packet can be calculated, resulting in multiple first values. Then, the arithmetic mean of these multiple first values ​​can be calculated using the target number of target data packets. The first variance is determined using the arithmetic mean and the first values. Through this embodiment, calculating the variance of the target protocol field value across multiple data packets allows for the evaluation of the field's stability. The more concentrated the field value distribution and the smaller the variance, the higher the consistency and stability of the field in data transmission. Calculating the first variance can also help detect abnormal data packets or protocol fields. If the variance of a certain field suddenly increases, an anomaly may have occurred during data transmission, such as data errors, network interference, or malicious attacks. The system can handle such anomalies through further analysis or alarm mechanisms.

[0077] In an exemplary embodiment, determining the first variance based on the arithmetic mean and the first numerical value includes: determining a first difference between each of the first numerical values ​​and the arithmetic mean; determining the square of each of the first differences to obtain a first squared value; determining the sum of the first squared values ​​to obtain a second sum; determining a fourth reciprocal of the second difference between the target quantity and a first constant; determining the arithmetic square root of the first product of the fourth reciprocal and the second sum as a first index; determining a second product of the first index, a first adaptive parameter, and a second constant; and determining the second power of the natural constant as the first variance.

[0078] In the above embodiment, the field dispersion index D p (That is, the first indicator mentioned above) can be calculated using the following formula: Here, N can be understood as the target number of captured target data packets. Specifically, for a certain initial semantic tag (e.g., the temperature field), the value of that field (i.e., the first value mentioned above) can be extracted from each target data packet, which can generate a set {x1, x2, ..., x...}. N}, x i This can be understood as representing the value of a certain field in the i-th target data packet; This can be understood as the arithmetic average of the field across N data packets; This is the first difference mentioned above; That is, the first squared value mentioned above; That is, the second sum mentioned above; This is the fourth reciprocal mentioned above.

[0079] In the above embodiments, the network traffic state variance can be represented by the field dispersion index D. p Calculated using the following formula: Wherein, λ is the first adaptive parameter mentioned above, which can be set according to empirical values; -λD p That is, the second product mentioned above; This is the second power of the product of the aforementioned natural constants. Through this embodiment, the method of determining the first variance based on the arithmetic mean and the first value not only improves the accuracy and efficiency of protocol parsing but also enhances the robustness and intelligence of the system when facing various IoT devices and complex network environments.

[0080] In an exemplary embodiment, determining the second variance of the target protocol field based on the target physical state data includes: determining an observation period for observing the target physical state data, and dividing the observation period into multiple sub-observation period periods; performing the following operations for each sub-observation period period to determine the second square value corresponding to the sub-observation period period: determining a first change in the target physical state data under the sub-observation period period and a second change in the second value of the target protocol field under the sub-observation period period; determining a first norm of the first change and a second norm of the second change; determining a third product of a second adaptive parameter and the second norm; determining a third difference between the first norm and the third product; determining the square of the third difference to obtain the second square value; determining the sum of the second square values ​​to obtain a third sum; determining a fourth ratio of the third sum to the total number of sub-observation period periods; determining the arithmetic square root of the fourth ratio as a second index; determining a fourth sum of the second index and a first constant; and determining a fifth ratio of the second index to the fourth sum as the second variance.

[0081] In the above embodiments, the second variance can be calculated from the physical synchronization residual Rs. The physical synchronization residual Rs (i.e., the aforementioned second index) can be understood as a measure of the degree of synchronization between protocol field changes and physical state changes based on device behavior (the smaller the residual, the better the synchronization). Physical synchronization residual Rs s It can be calculated using the following formula: Where k can be understood as the number of sub-observation periods into which the entire observation period is divided (i.e., the total number mentioned above), Δs k This can be understood as the change in the physical state of the equipment during the k-th time period (i.e., the first change mentioned above), ||Δs k || is the first norm mentioned above; Δx k This can be understood as the change in the protocol field value during the k-th time period (i.e., the second change mentioned above), ||Δx k || is the second norm mentioned above; β (the second adaptation parameter mentioned above) can be understood as a scaling factor, which can be used to map the changes in the protocol field to the scale of changes in the physical state. It can be estimated through methods such as linear regression. In ideal cases, s k =βx k β||Δx k || is the third product mentioned above; (||Δs) k ||-β||Δx k ||) is the third difference mentioned above; (||Δs) k ||-β||Δx k ||) 2That is, the second squared value mentioned above; That is, the third sum mentioned above; This is the fourth ratio mentioned above.

[0082] In the above embodiments, after determining the physical synchronization residual R s Then, the variance of the device behavior (i.e., the second variance mentioned above) can be calculated using the following formula: Among them, 1+R s This is the fourth sum mentioned above; This is the fifth ratio mentioned above. Through this embodiment, the method for determining the second variance of the target protocol field based on the target physical state data can achieve precise matching between the device's physical state and the communication protocol field. This not only improves the accuracy and efficiency of protocol parsing but also enhances the system's robustness and adaptability in the face of complex physical environments.

[0083] In one exemplary embodiment, determining the third difference of the target protocol field based on the target operation log includes: determining a first number of times the target device receives a target operation instruction based on the target operation log; determining a second number of times the target device responds to the target operation instruction; and determining the third difference based on the first number and the second number.

[0084] In the above embodiments, the third difference can be determined by the causal deviation C. d The causal deviation Cd can be calculated from the first and second counts. The causal deviation Cd can be understood as a measure of the stability of the causal relationship between the operation command and the device response, based on the user operation log (the smaller the deviation, the more stable the causal relationship). First, the first number of times the target device received the target operation command and the second number of times the target device responded to the target operation command can be determined from the target operation log. This immediately establishes the matching ratio of the response to the target operation command. Then, the third deviation can be determined using the first and second counts.

[0085] In this embodiment, calculating the third-party difference helps identify which protocol fields are closely related to user operation commands, thereby improving the parsing system's accuracy in recognizing and parsing control commands and ensuring the accuracy and timeliness of device responses. By observing the changes in the third-party difference over time, parsing parameters can be automatically adjusted, and parsing strategies optimized.

[0086] In an exemplary embodiment, determining the third difference based on the first number and the second number includes: determining a sixth ratio of the second number to the first number; determining a fourth difference between the indicator function of each target operation instruction and the sixth ratio; determining the square of each fourth difference to obtain a third square value; determining a fifth sum of each third square value; determining the arithmetic square root of the seventh ratio of the first constant to the fifth sum as a third index; and determining the hyperbolic tangent function value of the fourth product of the third index and the third adaptive parameter as the third difference.

[0087] In the above embodiments, the third difference can be calculated using the causal deviation Cd (i.e., the aforementioned third index), which can be calculated using the following formula: Where M can be understood as the number of user operation commands identified from the user operation log; o m This can be understood as the m-th operation instruction (i.e., each of the above target operation instructions); This can be understood as an indicator function. When the device response (originating from data packets returned by the device in network traffic or changes in device status) after the m-th operation command meets expectations, the value is 1; otherwise, it is 0. global (i.e., the sixth ratio mentioned above) can be understood as the global average matching rate (matching judgment: the expected response of the operation command needs to be defined in advance, such as opening the switch and receiving the switch response packet), that is, the proportion of responses that match among all operation commands (the ratio of the second number to the first number). This is the fourth difference mentioned above; That is, the third squared value mentioned above; This is the seventh ratio mentioned above.

[0088] In the above embodiments, after determining the third indicator, the third difference can be determined from the third indicator using the following formula: Here, γ is the third adaptive parameter mentioned above, which can be configured according to the target device type; γC d This is the fourth product mentioned above; tanh(γC) d This is the value of the hyperbolic tangent function mentioned above.

[0089] In the above embodiments, D p You can filter protocol dynamic fields, R s The physical mapping field can be locked, C d Control fields can be identified. Through this embodiment, the calculation of third-party differences can quantify the consistency between operation commands and device responses, thereby assisting in evaluating the stability and reliability of protocol fields in device control.

[0090] In an exemplary embodiment, determining the target semantic label of the target protocol field from a plurality of initial semantic labels based on the target weight includes: performing the following operations for each initial semantic label to determine a comparison result corresponding to the initial semantic label: determining a target filtering strategy based on the target weight; determining target data based on the target filtering strategy, wherein the target data includes the target original data, the target physical state data, and data included in the target operation log; determining a third value corresponding to the target data; and determining a fourth value of the initial semantic label stored in the target database; comparing the third value with the fourth value to obtain the comparison result; determining candidate semantic labels based on the comparison result; and determining the target semantic label based on the candidate semantic labels.

[0091] In the above embodiments, after determining the target weight, the initial semantic fields of the target protocol fields can be initially filtered using the target weight. Figure 10 This is a flowchart of the protocol dynamic field filtering process according to an embodiment of this application, such as... Figure 10 As shown, the default setting can be device behavior priority. Then, different target filtering strategies can be determined by comparing the target weight with the threshold. Different target data can be determined based on different target filtering strategies to achieve initial screening. The target filtering strategies can include device behavior-driven, operation command-driven, and network traffic-driven.

[0092] In the above embodiment, after selecting the target data, the third value of the field in the target data can be extracted. This third value is then compared with the fourth value of the initial semantic label included in the typical feature library (i.e., the target database mentioned above), yielding a comparison result. Using this comparison result, mismatched semantic labels in the initial semantic label set are removed to obtain candidate semantic labels. These candidate semantic labels can then be further filtered (i.e., filtered using multimodal joint processing) to obtain the target semantic label. This embodiment allows for preliminary filtering using data that plays a dominant role in the target weight, fully considering the influence of larger target weight values. Furthermore, setting an initial screening effectively simplifies the computational workload of subsequent fine-tuning and improves accuracy.

[0093] In an exemplary embodiment, determining a candidate semantic label based on the comparison result includes: determining a target comparison result included in the comparison result, wherein the target comparison result indicates that the third value is the same as the fourth value; and determining the initial semantic label corresponding to the target comparison result as the candidate semantic label.

[0094] In the above embodiments, a target comparison result can be selected from the comparison results where the third value is the same as or the difference is within a certain range. The semantic label corresponding to this target comparison result is determined as a candidate semantic label, which means that other semantic labels with large differences can be removed from the initial semantic labels. For example, if we want to determine whether the target protocol field is a temperature field, and the third value of this target protocol field is 50, while the fourth value of the temperature field stored in the target database is 45-55, then the initial semantic label "temperature" can be determined as a candidate semantic label, and multimodal processing can be used for further analysis. Through this embodiment, when the third value and the fourth value are the same or have a small difference, it can be indicated that the selected semantic label completely matches the actual data characteristics, which can greatly improve the accuracy of the parsing results and reduce equipment misoperation or data analysis deviations caused by parsing errors.

[0095] In one exemplary embodiment, determining a target filtering strategy based on the target weight includes: if a second target sub-weight included in the target weight is greater than a first preset threshold, determining the target filtering strategy as a device behavior-driven strategy, wherein the second target sub-weight is the weight of the target physical state data; if the second target sub-weight included in the target weight is less than the first preset threshold and a third target sub-weight is greater than the second preset threshold, determining the target filtering strategy as an operation instruction-driven strategy, wherein the third target sub-weight is the weight of the target operation log; if the second target sub-weight included in the target weight is less than the first preset threshold and the third target sub-weight is less than the second preset threshold, determining the target filtering strategy as a network traffic-driven strategy.

[0096] In the above embodiments, see again Figure 10 The second target sub-weight w is calculated using the target physical state data. s If the value is greater than the first preset threshold of 0.7, the target selection strategy can be determined to be device behavior-driven; in the second target sub-weight w s The second target sub-weight w is less than the first preset threshold of 0.7 and determined by the target operation log. c If the target weight is greater than the second preset threshold of 0.6, the target selection strategy can be determined as operation instruction guidance; in the case of the second target sub-weight w cWhen the threshold value is less than 0.6 (the second preset threshold), the target filtering strategy can be determined as a network traffic-driven strategy. Processes 1-4 in the diagram can be obtained by processing field bytes and corresponding modal information, corresponding to different target filtering strategies. The first and second preset thresholds can be adjusted according to actual conditions, and this invention does not impose any limitations on this. By assigning different modal weight threshold judgment conditions, differentiated field positioning strategies can be implemented. Furthermore, multimodal joint filtering can be added as a fallback strategy to maximize accuracy.

[0097] In one exemplary embodiment, determining target data based on the target filtering strategy includes: determining the target data as target physical state data when the target filtering strategy is a device behavior-driven strategy; determining the target data as target operation logs when the target filtering strategy is an operation instruction-driven strategy; and determining the target data as target raw data when the target filtering strategy is a network traffic-driven strategy.

[0098] In the above embodiments, when the target filtering strategy is a device behavior-driven strategy, it can be understood as comparing target physical state data related to the device behavior-driven strategy, i.e., the target data is the target physical state data; when the target filtering strategy is an operation command-driven strategy, it can be understood as comparing target operation logs related to the operation command-driven strategy, i.e., the target data is the target operation logs; when the target filtering strategy is a network traffic-driven strategy, it can be understood as comparing target raw data related to the network traffic-driven strategy, i.e., the target data is the target raw data. By determining the target filtering strategy, the dominant data in the multimodal data can be identified. By comparing this data with the reference values ​​stored in the target database, a preliminary filtering of the initial semantic tags can be performed, thereby improving the accuracy of the parsing results.

[0099] In an exemplary embodiment, determining the target semantic label based on the candidate semantic labels includes: performing the following operations for each candidate semantic label to determine a target confidence level for the candidate semantic label; determining the target semantic label based on the target confidence level: determining candidate raw data, candidate physical state data, and candidate operation logs corresponding to each candidate semantic label; determining a first objective function based on the candidate raw data, the first objective function representing the strength of evidence that the target protocol field is labeled as the candidate semantic label in a first modality; determining a second objective function based on the candidate physical state data, the second objective function representing the strength of evidence that the target protocol field is labeled as the candidate semantic label in a second modality; determining a third objective function based on the candidate operation logs, the third objective function representing the strength of evidence that the target protocol field is labeled as the candidate semantic label in a third modality; and determining the target confidence level based on the first objective function, the second objective function, and the third objective function.

[0100] In the above embodiments, the candidate semantic tags have been screened out through preliminary screening, so the unique real semantic field, i.e. the target semantic tag, can be determined from the candidate semantic tags.

[0101] In the above embodiments, a target confidence score can be calculated for each semantic label in the candidate semantic labels, and the target semantic label can be determined by comparing the target confidence scores. Candidate original data corresponding to the candidate semantic labels can be determined from the target original data, candidate physical state data corresponding to the candidate semantic labels can be determined from the target physical state data, and candidate operation logs corresponding to the candidate semantic labels can be determined from the target operation logs. Using the candidate original data, a first objective function representing the strength of evidence that the target protocol field is labeled as a candidate semantic label in the first modality (network traffic modality) can be determined. Similarly, using the candidate physical state data, a second objective function representing the strength of evidence that the target protocol field is labeled as a candidate semantic label in the second modality (device behavior modality) can be determined. Finally, using the candidate operation logs, a third objective function representing the strength of evidence that the target protocol field is labeled as a candidate semantic label in the third modality (user operation modality) can be determined. Finally, the target confidence score can be determined using the first, second, and third objective functions.

[0102] This embodiment demonstrates how the process of determining the target confidence level of each candidate semantic label to ultimately select the target semantic label can quantify the degree of matching between each candidate label and the actual data in different modalities, thereby improving the accuracy and reliability of semantic parsing.

[0103] In an exemplary embodiment, determining the target confidence level based on the first objective function, the second objective function, and the third objective function includes: determining a fifth product of a first target sub-weight included in the target weights and the first objective function; determining a sixth product of a second target sub-weight included in the target weights and the second objective function; determining a seventh product of a third target sub-weight included in the target weights and the third objective function; and determining the sixth sum of the fifth product, the sixth product, and the seventh product as the target confidence level.

[0104] In the above embodiments, the target confidence score can be calculated using the following formula: Confidenct(l,f i )=∑ m∈{p,s,c} w m *C m (l,f i ), where m represents the three modes mentioned above, namely network traffic mode p, device behavior mode s, and operation command mode c; w m Including w p (i.e., the first objective sub-weights mentioned above), w s (i.e., the second objective sub-weights mentioned above) and w c (i.e., the third objective sub-weights mentioned above); l can represent different candidate semantic labels (e.g., temperature, humidity, on / off status, etc.); f i This can be understood as a candidate field in the protocol, that is, the field being attempted to be labeled (target protocol field); C m (l,f i This can be understood as a strength of evidence function, that is, under modality m, field f i The strength of evidence labeled as tag l, with a value range of [0,1]. A larger value indicates a greater likelihood that the field is labeled l. Therefore, C... m (l,f i ) can include C p (l,f i (i.e., the first objective function mentioned above), C s (l,f i (i.e., the second objective function mentioned above) and C c (l,f i (i.e., the third objective function mentioned above); w p *C p (l,f i (that is, the fifth product mentioned above), w s *C s (l,f i (that is, the sixth product mentioned above), w c *C c (l,f i(That is, the seventh product mentioned above); ∑ m∈{p,s,c} w m *C m (l,f i This is the sixth sum mentioned above.

[0105] In this embodiment, the calculation of target confidence integrates data information from three modalities: network traffic, device behavior, and user operations. This ensures the comprehensiveness and accuracy of the decision-making process and avoids analytical errors caused by incomplete information from a single modality. By multiplying the target sub-weights by the objective function, the comparison weights of different modalities can be automatically adjusted according to specific scenarios, ensuring that the analytical results better match actual needs and improving the intelligence and flexibility of analytical decision-making.

[0106] In an exemplary embodiment, determining a first objective function based on the candidate raw data and the candidate semantic labels includes: determining a first feature of the candidate raw data under the candidate semantic labels and a second feature of the candidate semantic labels in a target database; determining a fifth difference between the first feature and the second feature; determining the square of the fifth difference to obtain a fourth squared value; determining the square of a predetermined deviation coefficient to obtain a fifth squared value; determining an eighth product of a third constant and the fifth squared value; determining an eighth ratio of the fifth difference to the eighth product; and determining the exponential function of the eighth ratio as the first objective function.

[0107] In the above embodiment, the first objective function C p (l,f i The degree of matching between the statistical characteristics of a field and a typical feature library of labels can be calculated, that is, determined through the statistical characteristics of the field and predefined features. p The lower the value, the larger the deviation between the value range of the target protocol field and the value range of the typical feature library.

[0108] In the above embodiments, C p (l,f i The calculation formula for ) is as follows: Among them, V field This can be understood as the statistical characteristics of the field (i.e., the first characteristic mentioned above, such as value range, byte distribution, etc.); V typical This can be understood as a predefined feature of the candidate semantic label in the target database (i.e., the second feature mentioned above, for example, the typical value range of a switch label is a Boolean value); σ (i.e., the deviation coefficient mentioned above) can be understood as the allowed deviation range, which can be configured according to the type value range; V field -V typical This is the fifth difference mentioned above; (V) field -V typical ) 2 This is the fourth squared value mentioned above; σ2 This is the fifth squared value mentioned above; This is the eighth ratio mentioned above; This is the exponential function of the eighth ratio mentioned above.

[0109] In this embodiment, the first objective function provides a method for quantifying the degree of feature matching by calculating an exponential function of the ratio of the square of the feature difference to the square of the deviation coefficient. This converts the degree of matching into a numerical value, which facilitates comparison and decision-making in the analytical system.

[0110] In one exemplary embodiment, based on the candidate physical state data and determining a second objective function, the method includes: determining a first change sequence of the target protocol field at adjacent time points; determining a second change sequence of the candidate physical state data at adjacent time points; and determining the absolute value of the Pierre correlation coefficient of the first change sequence and the second change sequence as the second objective function.

[0111] In the above embodiments, the second objective function C s (l,f i The correlation coefficient between field changes and sensor changes can be calculated, as shown in the formula below: C s (l,f i )=|corr(Δf i ,Δs l )|, where Δf i This can be understood as field f i The change sequence at multiple time points (the difference between adjacent time points, i.e., the first change sequence mentioned above); Δs l This can be understood as the sequence of changes in the physical sensors corresponding to the candidate semantic labels at the same point in time (i.e., the second change sequence mentioned above); corr can be understood as the Pearson correlation coefficient, i.e., |corr(Δf) i ,Δs l )| is the absolute value of the Pierre correlation coefficient of the first and second variation sequences mentioned above.

[0112] In this embodiment, the second objective function can quantify the correlation between protocol field changes and physical state data changes by calculating the Pearson correlation coefficient of the change sequence. This can help identify protocol fields closely related to the physical state and improve the accuracy of the analysis.

[0113] In one exemplary embodiment, determining a third objective function based on the candidate operation log includes: determining a first number of times the target device receives a target operation instruction; determining a second number of times the target device responds to the target operation instruction; and determining a ninth ratio of the first number and the second number as the third objective function.

[0114] In the above embodiments, the third objective function C c (l,f i The frequency of field changes after user interaction can be calculated, i.e., it can be determined by the first and second counts, as shown in the following formula: The number of matching operations can be understood as the number of times the target user performs operations related to the candidate semantic tags, as shown in field f. i The number of times the expected change occurred (i.e., the number of the second time); the total number of operations can be understood here as the total number of operations performed by the target user (i.e., the number of the first time mentioned above); This is the ninth ratio mentioned above. Through this embodiment, the third objective function, by calculating the ratio of the number of received operation commands to the number of responses, can quantify the consistency of device operation responses, helping to identify whether the device has correctly responded to user commands and improving the reliability of the parsing process.

[0115] In an exemplary embodiment, determining the target semantic label based on the target confidence level includes: determining the candidate semantic label corresponding to the maximum confidence level included in the target confidence level as the target semantic label; or, determining a sixth difference between each pair of target confidence levels, and if the sixth difference includes a seventh difference less than a third preset threshold, determining the target weight corresponding to the seventh difference, and determining the candidate semantic label corresponding to the maximum target weight as the target semantic label; or, if the sixth difference includes a seventh difference less than a third preset threshold, determining the similarity between the candidate semantic labels corresponding to the target confidence level of the seventh difference, merging the candidate semantic labels with similarity greater than a fourth preset threshold to obtain a merged semantic label, and determining the merged semantic label as the target semantic label.

[0116] In the above embodiment, the evidence strength function (i.e., the first objective function C) under the three modalities is used. p The second objective function C s and the third objective function C c According to the first objective sub-weight w p Second objective sub-weight C s and the third objective function w c Add them together, that is, Confidence = w p ·C p +w s ·C s +w c ·C c This allows us to obtain the target confidence level that is located between [0,1] after fusion.

[0117] In the above embodiments, after calculating the target confidence score for each candidate semantic label, the candidate semantic label corresponding to the highest target confidence score (i.e., the highest confidence score mentioned above) can be selected as the target semantic label.

[0118] In the above embodiments, if there is a situation where the seventh difference between the target confidence scores is less than the third threshold (i.e., the confidence scores of two candidate labels are very close and a conflict occurs), then the conflict resolution mechanism can be entered: firstly, it can be determined whether there is a situation where the weight of a modality is particularly high (e.g., greater than 0.7, this value is only an illustrative example and the present invention does not limit it), and the candidate semantic label supported by that modality can be used as the target semantic label.

[0119] In the above embodiments, if no modality has a particularly high weight, the semantic similarity between candidate semantic tags corresponding to the seventh difference can be calculated (e.g., through word vectors or knowledge graphs). If the similarity is higher than the fourth preset threshold, the candidate semantic tags can be merged (e.g., 'temperature' and 'water temperature' can be merged into 'temperature'), and the merged semantic tag is determined as the target semantic tag. If none of the above situations exist, manual intervention can be triggered. Through this embodiment, multimodal heterogeneous data from the Internet of Things can be transformed into quantifiable decision-making basis, which can effectively solve the pain point of protocol parsing but semantic agnosticness.

[0120] In an exemplary embodiment, before acquiring the raw data, physical status data, and operation log transmitted by the target device, the method further includes: acquiring the raw byte stream of the target device; removing the header information and metadata of the data packets included in the raw byte stream to obtain a standardized byte stream; and injecting target metadata into the standardized byte stream to obtain the raw data, wherein the target metadata is known protocol metadata.

[0121] In the above embodiments, for the raw byte stream transmitted by the target device, unified payload extraction and protocol feature injection processing are also required. After receiving the raw byte stream, the protocol parsing packet engine can automatically identify the transmission protocol and remove the header information and metadata of the data packets included in the raw byte stream to obtain a standardized byte stream, as shown in Table 3. If it is TCP / UDP (User Datagram Protocol), the IP (Internet Protocol) / TCP header can be stripped to extract the payload; if it is MQTT, the PUBLISH message can be parsed to extract the payload; if it is CoAP, the Option can be parsed to extract the payload; if it is HTTP (Hypertext Transfer Protocol), the Body part can be parsed. The data communication methods include, but are not limited to, MQTT, WEBSOCKET, TCP, HTTP, MODBUS (Modbus Communication Protocol), DALI (Digital Addressable Lighting Interface), and KNX (Konnex Association). The extracted standardized byte stream can also be injected with protocol metadata (i.e., the target metadata mentioned above) during the semantic annotation stage. For example, the metadata for the MQTT protocol is: {"protocol_meta":{"transport":"MQTT","topic":" / sensor / uplink / light","uri:""},"fields":[{"name:"switch","position":[0,1]}]}.

[0122] Table 3

[0123]

[0124] In the above embodiments, the automatic identification of transmission protocols can be achieved through a combination of message header features and communication mode statistical analysis. This can be achieved through protocol fingerprint database matching and statistical protocol feature identification. Protocol fingerprint database matching can be understood as analyzing data packet features based on accumulated IoT device parsing protocols. For example, if the first two bits of a data packet are '\x10', it matches the MQTT Connect message features. Statistical protocol feature identification can be understood as cases where the frequency of occurrence of the lower four bits of the MQTT message header being 0001 is greater than 80%.

[0125] In this embodiment, the generation of standardized byte streams removes redundant information from data packets, such as flag bits and address information. This not only reduces the complexity of data processing but also accelerates the parsing process and reduces latency. Injecting target metadata, such as known protocol types and field meanings, into the data stream provides the parser with necessary contextual information. This enables the parser to perform more accurate field location and meaning inference based on the injected metadata when facing unknown or non-standard protocol data, thereby improving the accuracy and robustness of protocol parsing.

[0126] The platform access method in this application will be explained below with reference to specific embodiments:

[0127] Figure 11 This is a flowchart illustrating the platform access method according to a specific embodiment of this application, such as... Figure 11 As shown, the physical layer interface can decode signals transmitted by IoT devices, receive the raw bit stream from the devices, and perform clock synchronization. By connecting additional sensors to the IoT devices, it can acquire the physical state of the devices and perform timing alignment, such as digital quantities of current or temperature, or the on / off status of lights. User control commands are recorded through log point marking and screen recording, and the operation logs are structured. The protocol parsing packet can automatically identify the transmission protocol, strip the TCP / UDP headers, and extract protocol-specific metadata, such as extracting the Topic for MQTT, parsing the URI path for CoAP, and separating the Header and Body for HTTP. A unified payload handles length decoding, unifies byte order, and then outputs a standardized byte stream. Traffic modalities are obtained through field statistical feature extraction, behavioral modalities through physical quantity-byte stream correlation analysis, and operational modalities through command-response causal modeling. Finally, semantic fields (i.e., the aforementioned target semantic tags) are obtained through multimodal fusion calculation formulas and conflict arbitration. Based on the automatically labeled semantic fields, a protocol description file (DSL) that can be processed by a computer can be generated. The protocol template rendering engine parses the protocol description file and generates driver code that can be loaded and executed by the IoT platform. After the driver test case set of the target device is tested through the automated integration platform, it can be dynamically published to the IoT platform. The target device's connection server address is then changed to the IoT platform address, and the platform can receive the data reported by the target device.

[0128] This optional example demonstrates how trimodal fusion (network traffic + device behavior + operation commands) can resolve semantic ambiguities. Furthermore, it supports, but is not limited to, any byte-stream-based application layer protocol such as TCP / UDP / MQTT / CoAP / MODBUS / WEBSOCKET. Moreover, closed-loop verification can achieve driver accuracy >95%, while related technologies achieve <80%.

[0129] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0130] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / random access memory (RAM), magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0131] According to another aspect of the embodiments of this application, a platform access device is also provided. This platform access device can be used to implement the platform access method provided in the above embodiments, and details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0132] Figure 12 This is a structural block diagram of an optional platform access device according to an embodiment of this application, such as... Figure 12 As shown, the access device for the platform includes:

[0133] The acquisition module 1202 is used to acquire the raw data transmitted by the target device, physical status data, and operation logs.

[0134] The first determining module 1204 is used to determine the target protocol field included in the original data;

[0135] Matching module 1206 is used to match multiple initial semantic tags for the target protocol field;

[0136] The second determining module 1208 is used to determine, based on each initial semantic tag, target original data associated with the initial semantic tag from the original data, target physical state data associated with the initial semantic tag from the physical state data, and target operation log associated with the initial semantic tag from the operation log;

[0137] The third determining module 1210 is used to determine the target weight of the target protocol field under each initial semantic tag based on the target original data, the target physical state data and the target operation log, and to determine the target semantic tag of the target protocol field from multiple initial semantic tags based on the target weight;

[0138] Access module 1212 is used to connect the target device to the target platform based on the target semantic tag.

[0139] In an exemplary embodiment, the third determining module 1210 can determine the target weight of the target protocol field under each initial semantic tag based on the target original data, target physical state data, and the target operation log in the following manner: For any initial semantic tag, the following operations are performed to obtain the target weight of the initial semantic tag: Under the initial semantic tag, a first variance of the target protocol field is determined based on the target original data, a second variance of the target protocol field is determined based on the target physical state data, and a third variance of the target protocol field is determined based on the target operation log; the target weight is determined based on the first variance, the second variance, and the third variance.

[0140] In an exemplary embodiment, the third determining module 1210 may determine the target weight based on the first variance, the second variance, and the third variance in the following manner: determining a first reciprocal of the first variance, a second reciprocal of the second variance, and a third reciprocal of the third variance; determining a first sum of the first reciprocal, the second reciprocal, and the third reciprocal; determining a first ratio of the first reciprocal to the first sum as a first target sub-weight, a second ratio of the second reciprocal to the first sum as a second target sub-weight, and a third ratio of the third reciprocal to the first sum as a third target sub-weight; and determining the first target sub-weight, the second target sub-weight, and the third target sub-weight as the target weight.

[0141] In an exemplary embodiment, the third determining module 1210 may determine the first variance of the target protocol field based on the target raw data in the following manner: splitting the target raw data into multiple target data packets; determining the target number of the target data packets, determining the value of the target protocol field included in each target data packet, and obtaining multiple first values; determining the arithmetic mean of the multiple first values; and determining the first variance based on the arithmetic mean and the first values.

[0142] In an exemplary embodiment, the third determining module 1210 may determine the first variance based on the arithmetic mean and the first value in the following manner: determining a first difference between each first value and the arithmetic mean; determining the square of each first difference to obtain a first squared value; determining the sum of the first squared values ​​to obtain a second sum; determining a fourth reciprocal of the second difference between the target quantity and the first constant; determining the arithmetic square root of the first product of the fourth reciprocal and the second sum as a first index; determining a second product of the first index, the first adaptive parameter, and the second constant; and determining the second power of the natural constant as the first variance.

[0143] In an exemplary embodiment, the third determining module 1210 can determine the second variance of the target protocol field based on the target physical state data in the following manner: determining the observation time period for observing the target physical state data, and dividing the observation time period into multiple sub-observation time periods; performing the following operations for each sub-observation time period to determine the second square value corresponding to the sub-observation time period: determining the first change in the target physical state data under the sub-observation time period and determining the second change in the second value of the target protocol field under the sub-observation time period, determining the first norm of the first change and the second norm of the second change, determining the third product of the second adaptive parameter and the second norm, determining the third difference between the first norm and the third product; determining the square of the third difference to obtain the second square value; determining the sum of the second square values ​​to obtain the third sum; determining the fourth ratio of the third sum to the total number of sub-observation time periods; determining the arithmetic square root of the fourth ratio as the second index; determining the fourth sum of the second index and the first constant; and determining the fifth ratio of the second index to the fourth sum as the second variance.

[0144] In an exemplary embodiment, the third determining module 1210 may determine the third difference of the target protocol field based on the target operation log in the following manner: determining the first number of times the target device receives the target operation instruction based on the target operation log; determining the second number of times the target device responds to the target operation instruction; and determining the third difference based on the first number and the second number.

[0145] In an exemplary embodiment, the third determining module 1210 may determine the third difference based on the first number and the second number by: determining a sixth ratio of the second number to the first number; determining a fourth difference between the indicator function of each target operation instruction and the sixth ratio; determining the square of each fourth difference to obtain a third square value; determining a fifth sum of each third square value; determining the arithmetic square root of the seventh ratio of the first constant to the fifth sum as a third index; and determining the hyperbolic tangent function value of the fourth product of the third index and the third adaptive parameter as the third difference.

[0146] In an exemplary embodiment, the third determining module 1210 can determine the target semantic label of the target protocol field from a plurality of initial semantic labels based on the target weight in the following manner: for each initial semantic label, perform the following operations to determine the comparison result corresponding to the initial semantic label: determine a target filtering strategy based on the target weight, determine target data based on the target filtering strategy, wherein the target data is the target original data, the target physical state data, and the data included in the target operation log, determine a third value corresponding to the target data, and determine a fourth value of the initial semantic label stored in the target database; compare the third value with the fourth value to obtain the comparison result; determine candidate semantic labels based on the comparison result; and determine the target semantic label based on the candidate semantic labels.

[0147] In an exemplary embodiment, the third determining module 1210 may determine candidate semantic labels based on the comparison result in the following manner: determining a target comparison result included in the comparison result, wherein the target comparison result indicates that the third value is the same as the fourth value; and determining the initial semantic label corresponding to the target comparison result as the candidate semantic label.

[0148] In an exemplary embodiment, the third determining module 1210 can determine the target filtering strategy based on the target weight in the following manner: when the second target sub-weight included in the target weight is greater than the first preset threshold, the target filtering strategy is determined to be a device behavior-driven strategy, wherein the second target sub-weight is the weight of the target physical state data; when the second target sub-weight included in the target weight is less than the first preset threshold and the third target sub-weight is greater than the second preset threshold, the target filtering strategy is determined to be an operation instruction-driven strategy, wherein the third target sub-weight is the weight of the target operation log; when the second target sub-weight included in the target weight is less than the first preset threshold and the third target sub-weight is less than the second preset threshold, the target filtering strategy is determined to be a network traffic-driven strategy.

[0149] In an exemplary embodiment, the third determining module 1210 can determine target data based on the target filtering strategy in the following ways: when the target filtering strategy is a device behavior-driven strategy, the target data is determined to be target physical state data; when the target filtering strategy is an operation instruction-driven strategy, the target data is determined to be target operation log; when the target filtering strategy is a network traffic-driven strategy, the target data is determined to be target raw data.

[0150] In an exemplary embodiment, the third determining module 1210 may determine the target semantic label based on the candidate semantic labels in the following manner: performing the following operations for each candidate semantic label to determine the target confidence level of the candidate semantic label, and determining the target semantic label based on the target confidence level: determining candidate raw data, candidate physical state data, and candidate operation logs corresponding to each candidate semantic label; determining a first objective function based on the candidate raw data, the first objective function being used to represent the evidence strength that the target protocol field is labeled as the candidate semantic label in a first modality; determining a second objective function based on the candidate physical state data, the second objective function being used to represent the evidence strength that the target protocol field is labeled as the candidate semantic label in a second modality; determining a third objective function based on the candidate operation logs, the third objective function being used to represent the evidence strength that the target protocol field is labeled as the candidate semantic label in a third modality; and determining the target confidence level based on the first objective function, the second objective function, and the third objective function.

[0151] In an exemplary embodiment, the third determining module 1210 can determine the target confidence level based on the first objective function, the second objective function, and the third objective function in the following manner: determining the fifth product of the first target sub-weights included in the target weights and the first objective function; determining the sixth product of the second target sub-weights included in the target weights and the second objective function; determining the seventh product of the third target sub-weights included in the target weights and the third objective function; and determining the sixth sum of the fifth product, the sixth product, and the seventh product as the target confidence level.

[0152] In an exemplary embodiment, the third determining module 1210 can determine a first objective function based on the candidate original data and the candidate semantic label in the following manner: determining a first feature of the candidate original data under the candidate semantic label and a second feature of the candidate semantic label in the target database; determining a fifth difference between the first feature and the second feature; determining the square of the fifth difference to obtain a fourth squared value; determining the square of a predetermined deviation coefficient to obtain a fifth squared value; determining an eighth product of a third constant and the fifth squared value; determining an eighth ratio of the fifth difference and the eighth product; and determining the exponential function of the eighth ratio as the first objective function.

[0153] In an exemplary embodiment, the third determining module 1210 may determine the second objective function based on the candidate physical state data by: determining a first change sequence of the target protocol field at adjacent time points; determining a second change sequence of the candidate physical state data at adjacent time points; and determining the absolute value of the Pierre correlation coefficient of the first change sequence and the second change sequence as the second objective function.

[0154] In an exemplary embodiment, the third determining module 1210 may determine a third objective function based on the candidate operation log by: determining a first number of times the target device receives a target operation instruction; determining a second number of times the target device responds to the target operation instruction; and determining a ninth ratio of the first number and the second number as the third objective function.

[0155] In an exemplary embodiment, the third determining module 1310 may determine the target semantic label based on the target confidence level in the following ways: determining the candidate semantic label corresponding to the maximum confidence level included in the target confidence level as the target semantic label; or, determining a sixth difference between each pair of target confidence levels, and if the sixth difference includes a seventh difference less than a third preset threshold, determining the target weight corresponding to the seventh difference, and determining the candidate semantic label corresponding to the maximum target weight as the target semantic label; or, if the sixth difference includes a seventh difference less than a third preset threshold, determining the similarity of the candidate semantic labels corresponding to the target confidence level corresponding to the seventh difference, merging the candidate semantic labels with similarity greater than a fourth preset threshold to obtain a merged semantic label, and determining the merged semantic label as the target semantic label.

[0156] In an exemplary embodiment, the apparatus is further configured to: acquire the original byte stream of the target device before acquiring the original data, physical status data, and operation log transmitted by the target device; remove the header information and metadata of the data packets included in the original byte stream to obtain a standardized byte stream; inject the target metadata into the standardized byte stream to obtain the original data, wherein the target metadata is known protocol metadata.

[0157] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0158] According to another aspect of the embodiments of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein the program executes the steps in any of the above method embodiments when it is run.

[0159] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, ROMs, RAMs, portable hard drives, magnetic disks, or optical disks.

[0160] According to another aspect of the embodiments of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor is configured to perform the steps of any of the method embodiments described above via the computer program. In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0161] Embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the methods in various embodiments of the present application.

[0162] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0163] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0164] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. An access method of a platform, characterized by, include: Acquire the raw data, physical status data, and operation logs transmitted by the target device; Determine the target protocol fields included in the raw data; Match multiple initial semantic tags for the target protocol field; Based on each initial semantic tag, target original data associated with the initial semantic tag is determined from the original data, target physical state data associated with the initial semantic tag is determined from the physical state data, and target operation log associated with the initial semantic tag is determined from the operation log; Based on the target raw data, the target physical state data, and the target operation log, the target weight of the target protocol field under each initial semantic tag is determined, and the target semantic tag of the target protocol field is determined from multiple initial semantic tags based on the target weight; Determining the target weight of the target protocol field under each initial semantic tag based on the target raw data, target physical state data, and the target operation log includes: performing the following operations for any initial semantic tag to obtain the target weight of the initial semantic tag: under the initial semantic tag, determining a first variance of the target protocol field based on the target raw data, a second variance of the target protocol field based on the target physical state data, and a third variance of the target protocol field based on the target operation log; determining the target weight based on the first variance, the second variance, and the third variance; and determining the target weight based on the first variance, the second variance, and the third variance. Determining the target weight using variance, the second variance, and the third variance includes: determining a first reciprocal of the first variance, a second reciprocal of the second variance, and a third reciprocal of the third variance; determining a first sum of the first reciprocal, the second reciprocal, and the third reciprocal; determining a first ratio of the first reciprocal to the first sum as a first target sub-weight, a second ratio of the second reciprocal to the first sum as a second target sub-weight, and a third ratio of the third reciprocal to the first sum as a third target sub-weight; and determining the first target sub-weight, the second target sub-weight, and the third target sub-weight as the target weight. The target device is connected to the target platform based on the target semantic tag.

2. The method of claim 1, wherein, Determining the first variance of the target protocol field based on the target raw data includes: The original target data is split into multiple target data packets; Determine the target number of the target data packets; Determine the value of the target protocol field included in each target data packet to obtain multiple first values; Determine the arithmetic mean of multiple first values; The first variance is determined based on the arithmetic mean and the first value.

3. The method according to claim 2, characterized in that, Determining the first variance based on the arithmetic mean and the first value includes: Determine a first difference between each of the first values ​​and the arithmetic mean; Determine the square of each of the first differences to obtain the first squared value; Determine the sum of the first squared values ​​to obtain the second sum; Determine the fourth reciprocal of the second difference between the target quantity and the first constant; The arithmetic square root of the first product of the fourth reciprocal and the second sum is determined as the first index; Determine the second product of the first index, the first adaptive parameter, and the second constant; The second power of the natural constant is determined as the first variance.

4. The method according to claim 1, characterized in that, Determining the second variance of the target protocol field based on the target physical state data includes: Determine the observation time period for observing the target's physical state data, and divide the observation time period into multiple sub-observation time periods; For each of the sub-observation time periods, the following operations are performed to determine the second squared value corresponding to the sub-observation time period: determining the first change in the target physical state data under the sub-observation time period and the second change in the second value of the target protocol field under the sub-observation time period; determining the first norm of the first change and the second norm of the second change; determining the third product of the second adaptive parameter and the second norm; determining the third difference between the first norm and the third product; and determining the square of the third difference to obtain the second squared value. Determine the sum of the second squared values ​​to obtain the third sum; Determine the fourth ratio of the third sum to the total number of the sub-observation time periods; The arithmetic square root of the fourth ratio is determined as the second index; Determine the fourth sum of the second index and the first constant; The second variance is determined by the fifth ratio of the second index to the fourth sum.

5. The method according to claim 1, characterized in that, Determining the third-party difference of the target protocol field based on the target operation log includes: Based on the target operation log, determine the first number of times the target device received the target operation command; Determine the second number of times the target device responds to the target operation command; The third difference is determined based on the first number and the second number.

6. The method according to claim 5, characterized in that, Determining the third difference based on the first number and the second number includes: Determine the sixth ratio of the second number to the first number; Determine the fourth difference between the indication function of each target operation instruction and the sixth ratio; The square of each of the fourth differences is determined to obtain the third square value; Determine the fifth sum value for each of the third squared values; The arithmetic square root of the seventh ratio of the first constant to the fifth sum is determined as the third index; The hyperbolic tangent function value of the fourth product of the third index and the third adaptive parameter is determined as the third difference.

7. The method according to claim 1, characterized in that, Determining the target semantic label of the target protocol field from a plurality of initial semantic labels based on the target weight includes: For each initial semantic tag, the following operations are performed to determine the comparison result corresponding to the initial semantic tag: determining a target filtering strategy based on the target weight, determining target data based on the target filtering strategy, wherein the target data includes the target original data, the target physical state data, and data included in the target operation log, determining a third value corresponding to the target data, and determining a fourth value of the initial semantic tag stored in the target database; comparing the third value with the fourth value to obtain the comparison result; Candidate semantic labels are determined based on the comparison results; The target semantic label is determined based on the candidate semantic labels.

8. The method according to claim 7, characterized in that, Based on the comparison results, candidate semantic tags are determined, including: Determine a target comparison result included in the comparison results, wherein the target comparison result indicates that the third value is the same as the fourth value; The initial semantic label corresponding to the target comparison result is determined as the candidate semantic label.

9. The method according to claim 7, characterized in that, Determining the target screening strategy based on the target weights includes: If the second target sub-weight included in the target weight is greater than the first preset threshold, the target screening strategy is determined to be a device behavior-driven strategy, wherein the second target sub-weight is the weight of the target physical state data; If the second target sub-weight included in the target weight is less than the first preset threshold and the third target sub-weight is greater than the second preset threshold, the target filtering strategy is determined to be an operation instruction-driven strategy, and the third target sub-weight is the weight of the target operation log. If the second target sub-weight included in the target weight is less than the first preset threshold and the third target sub-weight is less than the second preset threshold, the target screening strategy is determined to be a network traffic-driven strategy.

10. The method according to claim 7, characterized in that, Determining target data based on the aforementioned target filtering strategy includes: When the target selection strategy is a device behavior-driven strategy, the target data is determined to be target physical state data; When the target filtering strategy is an operation instruction-driven strategy, the target data is determined to be the target operation log; When the target filtering strategy is a network traffic-driven strategy, the target data is determined to be the target raw data.

11. The method according to claim 7, characterized in that, Determining the target semantic tag based on the candidate semantic tags includes: For each candidate semantic tag, the following operations are performed to determine the target confidence score of the candidate semantic tag, and to determine the target semantic tag based on the target confidence score: Determine the candidate original data, candidate physical state data, and candidate operation log corresponding to each candidate semantic label; A first objective function is determined based on the candidate raw data. The first objective function is used to represent the evidence strength that the target protocol field is labeled as the candidate semantic label in the first modality. A second objective function is determined based on the candidate physical state data. The second objective function is used to represent the evidence strength that the target protocol field is labeled as the candidate semantic label in the second modality. A third objective function is determined based on the candidate operation logs. The third objective function is used to represent the evidence strength that the target protocol field is labeled as the candidate semantic label in the third modality. The target confidence level is determined based on the first objective function, the second objective function, and the third objective function.

12. The method according to claim 11, characterized in that, Determining the target confidence level based on the first objective function, the second objective function, and the third objective function includes: Determine the fifth product of the first objective sub-weights included in the objective weights and the first objective function; Determine the sixth product of the second objective sub-weights included in the objective weights and the second objective function; Determine the seventh product of the third objective sub-weights included in the objective weights and the third objective function; The sixth sum of the fifth product, the sixth product, and the seventh product is determined as the target confidence level.

13. The method according to claim 11, characterized in that, Determining the first objective function based on the candidate raw data includes: Determine the first feature of the candidate original data under the candidate semantic label and the second feature of the candidate semantic label in the target database; Determine the fifth difference between the first feature and the second feature; The square of the fifth difference is determined to obtain the fourth square value; The fifth squared value is obtained by determining the square of the predetermined deviation coefficient; Determine the eighth product of the third constant and the fifth squared value; Determine the eighth ratio of the fifth difference to the eighth product; The exponential function of the eighth ratio is determined as the first objective function.

14. The method according to claim 11, characterized in that, Based on the candidate physical state data and the determination of the second objective function, including: Determine the first change sequence of the target protocol field at adjacent time points; Determine the second change sequence of the candidate physical state data at adjacent time points; The absolute values ​​of the Pierre correlation coefficients of the first and second variation sequences are used as the second objective function.

15. The method according to claim 11, characterized in that, The third objective function is determined based on the candidate operation logs, including: Determine the first number of times the target device receives the target operation command; Determine the second number of times the target device responds to the target operation command; The ninth ratio of the first number and the second number is determined as the third objective function.

16. The method according to claim 11, characterized in that, Determining the target semantic label based on the target confidence score includes: The candidate semantic label corresponding to the highest confidence level included in the target confidence level is determined as the target semantic label; or... Determine a sixth difference between each pair of target confidence scores. If the sixth difference includes a seventh difference less than a third preset threshold, determine the target weight corresponding to the seventh difference, and determine the candidate semantic label corresponding to the largest target weight as the target semantic label; or... If the sixth difference includes a seventh difference that is less than the third preset threshold, the similarity of the candidate semantic tags corresponding to the target confidence level corresponding to the seventh difference is determined, and the candidate semantic tags with similarity greater than the fourth preset threshold are merged to obtain a merged semantic tag, and the merged semantic tag is determined as the target semantic tag.

17. The method according to claim 1, characterized in that, Before acquiring the raw data, physical status data, and operation logs transmitted by the target device, the method further includes: Obtain the raw byte stream of the target device; The header information and metadata of the data packets included in the original byte stream are removed to obtain a standardized byte stream; The target metadata is injected into the standardized byte stream to obtain the original data, wherein the target metadata is known protocol metadata.

18. An access device for a platform, characterized in that, include: The acquisition module is used to acquire the raw data transmitted by the target device, physical status data, and operation logs. The first determining module is used to determine the target protocol field included in the original data; The matching module is used to match multiple initial semantic tags for the target protocol field; The second determining module is used to determine, based on each initial semantic tag, target original data associated with the initial semantic tag from the original data, target physical state data associated with the initial semantic tag from the physical state data, and target operation log associated with the initial semantic tag from the operation log; The third determining module is used to determine the target weight of the target protocol field under each initial semantic tag based on the target original data, the target physical state data and the target operation log, and to determine the target semantic tag of the target protocol field from multiple initial semantic tags based on the target weight; The third determining module determines the target weight of the target protocol field under each initial semantic tag based on the target original data, target physical state data, and the target operation log in the following manner: For any initial semantic tag, the following operations are performed to obtain the target weight of the initial semantic tag: Under the initial semantic tag, a first variance of the target protocol field is determined based on the target original data, a second variance of the target protocol field is determined based on the target physical state data, and a third variance of the target protocol field is determined based on the target operation log; the target weight is determined based on the first variance, the second variance, and the third variance. The third determining module determines the target weight based on the first variance, the second variance, and the third variance in the following manner: determining the first reciprocal of the first variance, the second reciprocal of the second variance, and the third reciprocal of the third variance; determining the first sum of the first reciprocal, the second reciprocal, and the third reciprocal; determining the first ratio of the first reciprocal to the first sum as the first target sub-weight, the second ratio of the second reciprocal to the first sum as the second target sub-weight, and the third ratio of the third reciprocal to the first sum as the third target sub-weight; and determining the first target sub-weight, the second target sub-weight, and the third target sub-weight as the target weight. The access module is used to connect the target device to the target platform based on the target semantic tag.

19. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 17.

20. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 17.

21. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 17.

Images