Time synchronization information security examination method, device and equipment based on dual-scale feature collaboration, and medium
By employing a dual-scale feature-coordinated time synchronization information security review method, which combines large-scale and small-scale branch neural networks, the problem of detecting both long-term and short-term attacks in time synchronization systems is solved, achieving efficient security review of time synchronization information and improving system security.
Patent Information
- Application Number
- CN202511823989.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-05
- Publication Date
- 2026-03-24
AI Technical Summary
Existing security audit methods for time synchronization systems struggle to effectively balance the multi-scale characteristics of long-term low-frequency attacks and short-term high-frequency attacks, resulting in insufficient accuracy and reliability in security threat detection.
A time-synchronized information security review method based on dual-scale feature collaboration is adopted. The method learns the long-term trend of time series through large-scale branch and captures short-term fluctuation features through small-scale branch. The dual optimization mechanism of residual sequence is used to realize cross-scale feature interaction and representation enhancement, and the final predicted value is generated for security review.
It significantly improves the accuracy and reliability of time synchronization information security review, effectively identifies long-term and short-term abnormal characteristics, enhances sensitivity to deviations from the baseline, and improves the security of the time synchronization system.
Smart Images

Figure CN121723339A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of time synchronization information security, and more specifically, to a time synchronization information security review method, apparatus, device, and medium based on dual-scale feature collaboration. Background Technology
[0002] As a core supporting technology of modern information infrastructure, time synchronization systems play an irreplaceable role in key fields such as power, communications and finance. However, time synchronization information faces security threats such as timestamp tampering attacks, delay attacks and replay attacks throughout its entire life cycle of generation, transmission, processing and use, becoming a weak link in the security of time synchronization systems.
[0003] Security audits of time-synchronized information not only need to identify systemic long-term benchmark drift and potential backdoor implantation, but also need to effectively detect malicious tampering and the subtle traces left in the time series. Any neglect or misjudgment of long-term security posture and short-term abnormal characteristics may lead to serious security omissions or misjudgments, weakening the credibility and practicality of the audit results.
[0004] However, existing security review methods face a prominent contradiction when conducting comprehensive time-series threat monitoring: if the model focuses on analyzing the long-term statistical regularities and stability characteristics of the time series, it may over-smooth high-frequency details and short-term fluctuations in the series. While it can identify systematic biases and low-frequency attacks, it struggles to capture rapidly changing attack behaviors. Conversely, if the model focuses on learning the short-term variation patterns and instantaneous fluctuations of the series, it will be affected by the noise accumulation effect of the time extrapolation process, easily overlooking long-latency, covert attacks, and even generating a large number of false alarms. The main reason why existing security review methods cannot effectively balance the above contradictions is that most security review models use static feature perception or fixed time-scale window modeling and analysis methods, failing to explicitly distinguish between the long-term and short-term characteristics of the time series, leading to information interference and making it difficult to effectively address the multi-scale characteristics of security threats in the time dimension. Summary of the Invention
[0005] To address the challenge of detecting multi-scale dynamic attacks in current time synchronization systems and to meet the security audit requirements for both long-term low-frequency attacks and short-term high-frequency attacks, this invention provides a time synchronization information security audit method, apparatus, device, and medium based on dual-scale feature collaboration. It is applicable to various time synchronization systems, such as NTP protocol, PTP protocol, and dual / unidirectional time synchronization systems.
[0006] In a first aspect, the present invention provides a time synchronization information security review method based on dual-scale feature collaboration, comprising: Generate large-scale sample sets based on historical time series; The large-scale branch in a dual-scale branch neural network model is trained using a large-scale sample set to generate a large-scale branch model. Small-scale sample sets are generated based on the predicted values of large-scale sample sets using a large-scale branching model. The small-scale branch in a dual-scale branch neural network model is trained using a small-scale sample set to generate a small-scale branch model. Generate final predicted values, which include the predictions of the large-scale branch model for the large-scale sample set and the predictions of the small-scale branch model for the small-scale sample set. Security audits are conducted on real-time time synchronization information based on the final predicted values.
[0007] Furthermore, the generation of a large-scale sample set based on historical time series includes: Obtain historical time series; Preprocess the historical time series; Large-scale sample sets are generated from the preprocessed historical time series using large-scale sampling and slicing strategies; The sliding shift operation is used to convert the training samples in the large-scale sample set into multiple sets of input-output tensor samples corresponding to the dual-scale branch neural network model.
[0008] Furthermore, the step of generating a small-scale sample set from the predicted values of the large-scale sample set based on the large-scale branching model includes: Large-scale branching models are used to predict large-scale sample sets, and the output is a predicted value that includes the long-term trend of the original time series. The residual sequence is obtained by subtracting the predicted values from the large-scale branch model from the original time series. Small-scale sample sets are generated from the residual sequences using small-scale sampling and slicing strategies. The sliding shift operation is used to convert the training samples in the small-scale sample set into multiple sets of input-output tensor samples corresponding to the dual-scale branch neural network model.
[0009] Further, generating the final predicted value includes: The small-scale branching model is used to predict small-scale sample sets, and the output is a predicted value that contains details of short-term fluctuations in the original time series. The predicted values output by the large-scale branch model and the small-scale branch model are superimposed to generate the final predicted value.
[0010] Furthermore, the security review of real-time time synchronization information based on the final predicted value includes: When the absolute difference between the final predicted value and the current time synchronization information exceeds a set threshold At that time, the review result was 1, indicating that the time synchronization information at the current moment posed a security threat; When the absolute difference between the final predicted value and the current time synchronization information is less than or equal to a set threshold If the review result is 0, the time synchronization information at the current moment is determined to be secure and reliable.
[0011] Furthermore, the time synchronization information security review method based on dual-scale feature collaboration also includes updating the local clock of the time synchronization system according to the security review results: If there is a security threat to the current time synchronization information, the local clock is updated using the final predicted value; If the current time synchronization information is secure and reliable, the local clock is updated using the current time synchronization information.
[0012] Secondly, the present invention provides a time-synchronized information security review device based on dual-scale feature collaboration, comprising: A dual-scale sequence construction module is used to generate large-scale sample sets based on historical time series; and to generate small-scale sample sets based on the predictions of the large-scale sample sets by the large-scale branch model. The dual-scale collaborative training module is used to train the large-scale branch of a dual-scale branch neural network model using a large-scale sample set to generate a large-scale branch model; to train the small-scale branch of a dual-scale branch neural network model using a small-scale sample set to generate a small-scale branch model; and to generate a final prediction value, which includes the prediction value of the large-scale branch model for the large-scale sample set and the prediction value of the small-scale branch model for the small-scale sample set. The security review module is used to conduct security reviews of real-time time synchronization information based on the final predicted value.
[0013] Furthermore, the time-synchronized information security review device based on dual-scale feature collaboration also includes: The local clock update module is used to update the local clock of the time synchronization system based on the results of the security review.
[0014] Thirdly, the present invention provides an electronic device, comprising: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores instructions that can be executed by the at least one processor, and the at least one processor executes the instructions stored in the memory to perform the method described above.
[0015] Fourthly, the present invention provides a computer-readable storage medium for storing instructions that, when executed, cause the above-described method to be implemented.
[0016] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This invention is based on a dual-scale branch neural network model with a collaborative training mechanism. The large-scale branch learns the long-term trend patterns of time series data to establish a reliable system security baseline; the small-scale branch captures short-term fluctuation features, enhancing sensitivity to short-term anomalies deviating from the baseline. Simultaneously, it innovatively designs a dual optimization mechanism based on residual sequences: during the training phase, residual information is passed from the large-scale branch to the small-scale branch, achieving cross-scale feature interaction and representation enhancement; during the review phase, the short-term fluctuation features extracted by the small-scale branch are dynamically integrated into the system security baseline assessment of the large-scale branch in a compensatory manner, achieving complementary enhancement of long- and short-scale feature extraction of time synchronization information. This achieves a synergistic balance between long-term security situation stability monitoring and sensitive identification of short-term anomaly features, significantly improving the accuracy and reliability of time synchronization information security review. Attached Figure Description
[0017] Figure 1 The flowchart illustrates a time synchronization information security review method based on dual-scale feature collaboration, as provided in this embodiment of the invention.
[0018] Figure 2 This is a schematic diagram of a time-synchronized information security review device based on dual-scale feature collaboration, provided as an embodiment of the present invention.
[0019] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0022] Time synchronization information exhibits significant non-stationary, nonlinear, and multi-noise superposition characteristics. It includes both the inherent frequency drift and drift rate variations of atomic frequency standards and is affected by complex phase noise introduced by external environmental disturbances. Based on the characteristics of time synchronization systems and the challenge of detecting multi-scale dynamic attacks, this invention provides a time synchronization information security review method based on dual-scale feature collaboration, such as... Figure 1 As shown, it includes: S100 generates a large-scale sample set based on historical time series. S200 uses a large-scale sample set to train the large-scale branch in a dual-scale branch neural network model, generating a large-scale branch model. S300 generates a small-scale sample set based on the predicted values of a large-scale sample set using a large-scale branch model. S400 uses a small-scale sample set to train the small-scale branch in a dual-scale branch neural network model, generating a small-scale branch model. S500, Generate the final predicted value, which includes the prediction value of the large-scale branch model for the large-scale sample set and the prediction value of the small-scale branch model for the small-scale sample set. S600 performs security audits on real-time time synchronization information based on the final predicted values.
[0023] The following details the specific implementation of the aforementioned time-synchronized information security review method based on dual-scale feature collaboration.
[0024] S100, generating a large-scale sample set based on historical time series: S101, Obtain historical time series; the historical time series is obtained and stored in advance and called when needed.
[0025] S102, preprocessing the historical time series; the preprocessing in this embodiment of the invention is to perform a difference processing on the historical time series to highlight the rate of change and high-frequency details of the time series, and to complete the removal of coarse errors, so as to provide high-quality data input for the subsequent collaborative training of the dual-scale branch neural network model.
[0026] S103 generates a large-scale sample set from the preprocessed historical time series using large-scale sampling and slicing strategies. For example, the generated large-scale sample set Represented as:
[0027] in, L This represents the prediction step size of the large-scale branch in a two-scale branch neural network model, where the large-scale sample set is used. have L Subsequences ; v This represents the input window length of the two-scale branch neural network model. The correspondence between the input and output of the two-scale branch neural network model in each subsequence is expressed as follows:
[0028] S104 uses a sliding shift operation to convert the training samples in the large-scale sample set into multiple sets of input-output tensor samples corresponding to the dual-scale branch neural network model. For example, when the window moves one step to the right, the input-output tensor samples are as follows:
[0029] S200 uses a large-scale sample set to train the large-scale branch in a dual-scale branch neural network model, generating a large-scale branch model. This large-scale branch needs to be selected from models that are good at capturing long-term dependencies, such as linear regression, ARIMA (Autoregressive Integrated Moving Average Model), and LSTM (Long Short-Term Memory).
[0030] S300 generates a small-scale sample set based on the predicted values of a large-scale sample set using a large-scale branching model: S301, utilizing a large-scale branching model For large-scale sample sets Make predictions and output predicted values that include the long-term trends of the original time series. , represented as:
[0031] S302, subtract the predicted values from the large-scale branch model from the original time series to obtain the residual series. , represented as:
[0032] The residual sequence It contains short-term fluctuations and detailed information that are ignored by large-scale branch models.
[0033] S303 generates a small-scale sample set from the residual sequence using a small-scale sampling and slicing strategy. , represented as:
[0034] in, S This represents the prediction step size of the small-scale branch in a two-scale branch neural network model, where the small-scale sample set is used. haveS Subsequences.
[0035] S304, using a sliding shift operation, transforms the training samples in the small-scale sample set into multiple sets of input-output tensor samples corresponding to the dual-scale branch neural network model. In this embodiment of the invention, the small-scale sample set... In terms of construction logic, it is similar to large-scale sample sets Similarly, the same input window length is used for all cases. v The two methods, using sliding shift operations, generate multiple sets of input-output tensor samples. The main difference lies in the prediction step size of the smaller-scale branch in the two-scale branch neural network model. S Furthermore, the interval between sample data in its subsequence is also... S This enables the extraction and modeling of short-term fluctuation characteristics.
[0036] S400 uses a small-scale sample set to train the small-scale branch in a dual-scale branch neural network model, generating a small-scale branch model. This small-scale branch needs to be selected from models that are good at capturing short-term fluctuations, such as more complex neural network models.
[0037] S500, Generate the final predicted values, which include the predictions from the large-scale branch model for the large-scale sample set and the predictions from the small-scale branch model for the small-scale sample set; specifically: S501, utilizing a small-scale branching model For small-scale sample sets Make predictions and output predicted values that include details of short-term fluctuations in the original time series. , represented as:
[0038] S502 combines the predicted values from the large-scale branch model and the small-scale branch model to generate the final predicted value. , represented as:
[0039] It can be seen that the final predicted value It provides a small-scale prediction correction value on the basis of large-scale trend prediction, so that the output of the dual-scale branch neural network model has a more detailed fluctuation description.
[0040] S600 performs security audits on real-time time synchronization information based on the final predicted values: When the absolute difference between the final predicted value and the current time synchronization information exceeds a set threshold At that time, the review result was 1, indicating that the time synchronization information at the current moment posed a security threat; When the absolute difference between the final predicted value and the current time synchronization information is less than or equal to a set threshold If the review result is 0, the time synchronization information at the current moment is determined to be secure and reliable.
[0041] Furthermore, the time-synchronization information security review method based on dual-scale feature collaboration also includes: S700 updates the local clock of the time synchronization system based on the results of the security review: If there is a security threat to the current time synchronization information, the local clock is updated using the final predicted value; If the current time synchronization information is secure and reliable, the local clock is updated using the current time synchronization information.
[0042] The following is an example of a time server obtaining time synchronization information from a higher-level time source via the NTP protocol.
[0043] The time series between the local clock and the upstream time source is continuously collected at 1-minute sampling intervals. The dual-scale branch neural network model analyzes the historical time series of the past 48 hours, therefore the input time series length is 2880. The specific parameters of the dual-scale branch neural network model are shown in Table 1.
[0044] Table 1. Two-scale branch neural network model
[0045] The above-described time-synchronization information security review method based on dual-scale feature collaboration is implemented as follows: (1) Data preprocessing: The historical time series is differenced once, and the MAD (Median Absolute Deviation) method is used to remove gross errors, thereby obtaining the model input series. .
[0046] (2) Constructing a large-scale sample set: The large-scale branch model achieves long-term prediction of 1 hour, which means that the large-scale branch model needs to predict 60 data points in the future. Therefore, the large-scale sample set is constructed in a column-first manner. Multiple sets of input-output tensor samples can be generated through sliding shift operations.
[0047] (3) Large-scale branch training: The input tensor shape is (34, 60, 14), indicating a total of 34 samples, each with 60 time steps and 14 feature variables per step. The output tensor shape is (34, 60, 1), indicating a total of 34 samples, each with 60 time steps, and each step corresponding to a target value. After 2000 iterations of training, predicted values that have learned the long-term trend of the original time series are obtained. .
[0048] (4) Construct a small-scale sample set, and use the original time series and the predicted values output by the large-scale branch model to calculate the residual series containing short-term fluctuations and details ignored by the large-scale branch model. And construct a small-scale sample set in a column-first manner. Multiple sets of input-output tensor samples can be generated by using sliding shift operations.
[0049] (5) Small-scale branch training: The input tensor shape is (562, 5, 14), indicating a total of 562 samples, each with 5 time steps and 14 feature variables per step. The output tensor shape is (562, 5, 1), indicating a total of 562 samples, each with 5 time steps, and each step corresponding to a target value. After 2000 iterations of training, predicted values that have learned the details of short-term fluctuations in the original time series are obtained. .
[0050] (6) Optimize the compensation output: the final predicted value of the dual-scale branch neural network model .
[0051] (7) Security audit and local clock update: Calculate the current time synchronization information. Compared with the final predicted value The absolute difference between them, and the set threshold. Compare them.
[0052] Scenario A: Normal fluctuation scenario, due to slight network congestion, the current time synchronization information is affected. Compared with the final predicted value The difference is 50ns, that is Less than the threshold The review result was 0, indicating that the current time synchronization information is secure and reliable, and the local clock will be updated directly using the current time synchronization information.
[0053] Scenario B: Delay attack scenario, where the attacker injects malicious delays into the time synchronization information, causing the current time synchronization information to be disrupted. Compared with the final predicted value An abnormal deviation of 500 ns occurred between them, that is Greater than the threshold The review result is 1, indicating that there is a security threat to the current time synchronization information, and the local clock will be updated using the final predicted value.
[0054] Based on the same technological concept, such as Figure 2As shown, this embodiment of the invention also provides a time-synchronized information security review device based on dual-scale feature collaboration, comprising: A dual-scale sequence construction module is used to generate large-scale sample sets based on historical time series; and to generate small-scale sample sets based on the predictions of the large-scale sample sets by the large-scale branch model. The dual-scale collaborative training module is used to train the large-scale branch of a dual-scale branch neural network model using a large-scale sample set to generate a large-scale branch model; to train the small-scale branch of a dual-scale branch neural network model using a small-scale sample set to generate a small-scale branch model; and to generate a final prediction value, which includes the prediction value of the large-scale branch model for the large-scale sample set and the prediction value of the small-scale branch model for the small-scale sample set. The security review module is used to conduct security reviews of real-time time synchronization information based on the final predicted value.
[0055] Furthermore, the time-synchronized information security review device based on dual-scale feature collaboration also includes: The local clock update module is used to update the local clock of the time synchronization system based on the results of the security review.
[0056] The specific working principles of each functional module in the above-mentioned device can be referred to the description in the foregoing method embodiments, and will not be repeated here.
[0057] Based on the same technical concept, embodiments of the present invention also provide an electronic device that can implement the multimodal fusion forged image recognition method flow provided in the above embodiments of the present invention. In one embodiment, the electronic device may be a server, a terminal device, or other electronic device. Figure 3 As shown, the electronic device may include: At least one processor and a memory connected to the at least one processor. In this embodiment of the invention, the specific connection medium between the processor and the memory is not limited. Figure 3 The example used is the connection between the processor and memory via a bus. The bus... Figure 3 The connections between other components are indicated by thick lines and are for illustrative purposes only, not as limiting information. Buses can be divided into address buses, data buses, control buses, etc., but for ease of representation, [the specific bus type is not shown here]. Figure 3 The processor is represented by a single thick line, but this does not imply that there is only one bus or one type of bus. Alternatively, a processor can also be called a controller; there are no restrictions on the name.
[0058] In this embodiment of the invention, the memory stores instructions that can be executed by at least one processor. By executing the instructions stored in the memory, at least one processor can execute a multimodal fusion forgery image recognition method described above.
[0059] The processor is the control center of the device. It can connect to various parts of the control equipment through various interfaces and lines. By running or executing instructions stored in memory and calling data stored in memory, it can monitor the various functions and data processing of the device as a whole.
[0060] In an alternative design, the processor may include one or more processing units. The processor may integrate an application processor and a modem processor, wherein the application processor primarily handles the operating system, user interface, and applications, while the modem processor primarily handles wireless communication. It is understood that the modem processor may also not be integrated into the processor. In some embodiments, the processor and memory may be implemented on the same chip; in some embodiments, they may also be implemented separately on separate chips.
[0061] The processor can be a general-purpose processor, such as a CPU, digital signal processor, application-specific integrated circuit, field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the multimodal fusion forged image recognition method disclosed in the embodiments of this invention can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.
[0062] Memory, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory can include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. Memory is any other medium capable of carrying or storing desired program code in the form of instructions or data structures, and accessible by a computer, but is not limited thereto. In embodiments of the present invention, memory can also be a circuit or any other device capable of implementing storage functions, used to store program instructions and / or data.
[0063] By designing and programming the processor, the code corresponding to the multimodal fusion forged image recognition method described in the foregoing embodiments can be embedded into the chip, enabling the chip to execute the steps of the method described in the foregoing embodiments during runtime. How to design and program the processor is a technique well-known to those skilled in the art and will not be elaborated upon here.
[0064] Based on the same inventive concept, embodiments of the present invention also provide a storage medium storing computer instructions that, when executed on a computer, cause the computer to perform a multimodal fusion method for identifying forged images as described above.
[0065] In some alternative embodiments, the present invention also provides a multimodal fusion method for identifying forged images that can also be implemented as a program product comprising program code that, when the program product is run on a device, causes the control device to perform the steps of the multimodal fusion method for identifying forged images according to various exemplary embodiments of the present invention as described above.
[0066] It should be noted that although several units or sub-units of the apparatus have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the invention, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units. Furthermore, although the operation of the method of the invention is described in a specific order in the drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.
[0067] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0068] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a server, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0069] Program code for performing the operations of this invention can be written using any combination of one or more programming languages, including object-oriented programming languages such as Java and C++, as well as conventional procedural programming languages such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.
[0070] In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).
[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A time-synchronized information security review method based on dual-scale feature collaboration, characterized in that, include: Generate large-scale sample sets based on historical time series; The large-scale branch in a dual-scale branch neural network model is trained using a large-scale sample set to generate a large-scale branch model. Small-scale sample sets are generated based on the predicted values of large-scale sample sets using a large-scale branching model. The small-scale branch in a dual-scale branch neural network model is trained using a small-scale sample set to generate a small-scale branch model. Generate final predicted values, which include the predictions of the large-scale branch model for the large-scale sample set and the predictions of the small-scale branch model for the small-scale sample set. Security audits are conducted on real-time time synchronization information based on the final predicted values.
2. The time synchronization information security review method based on dual-scale feature collaboration according to claim 1, characterized in that, The generation of large-scale sample sets based on historical time series includes: Obtain historical time series; Preprocess the historical time series; Large-scale sample sets are generated from the preprocessed historical time series using large-scale sampling and slicing strategies; The sliding shift operation is used to convert the training samples in the large-scale sample set into multiple sets of input-output tensor samples corresponding to the dual-scale branch neural network model.
3. The time synchronization information security review method based on dual-scale feature collaboration according to claim 1, characterized in that, The process of generating a small-scale sample set from the predicted values of a large-scale sample set based on a large-scale branching model includes: Large-scale branching models are used to predict large-scale sample sets, and the output is a predicted value that includes the long-term trend of the original time series. The residual sequence is obtained by subtracting the predicted values from the large-scale branch model from the original time series. Small-scale sample sets are generated from the residual sequences using small-scale sampling and slicing strategies. The sliding shift operation is used to convert the training samples in the small-scale sample set into multiple sets of input-output tensor samples corresponding to the dual-scale branch neural network model.
4. The time synchronization information security review method based on dual-scale feature collaboration according to claim 3, characterized in that, The generation of the final predicted value includes: The small-scale branching model is used to predict small-scale sample sets, and the output is a predicted value that contains details of short-term fluctuations in the original time series. The predicted values output by the large-scale branch model and the small-scale branch model are superimposed to generate the final predicted value.
5. The time synchronization information security review method based on dual-scale feature collaboration according to claim 1, characterized in that, The security review of real-time time synchronization information based on the final predicted value includes: When the absolute difference between the final predicted value and the current time synchronization information exceeds a set threshold At that time, the review result was 1, indicating that the time synchronization information at the current moment posed a security threat; When the absolute difference between the final predicted value and the current time synchronization information is less than or equal to a set threshold If the review result is 0, the time synchronization information at the current moment is determined to be secure and reliable.
6. The time synchronization information security review method based on dual-scale feature collaboration according to claim 5, characterized in that, This also includes updating the local clock of the time synchronization system based on the results of security audits: If there is a security threat to the current time synchronization information, the local clock is updated using the final predicted value; If the current time synchronization information is secure and reliable, the local clock is updated using the current time synchronization information.
7. A time-synchronized information security review device based on dual-scale feature collaboration, characterized in that, include: A dual-scale sequence construction module is used to generate large-scale sample sets based on historical time series. And for generating small-scale sample sets from predictions of large-scale sample sets based on large-scale branching models; The dual-scale collaborative training module is used to train the large-scale branch in the dual-scale branch neural network model using a large-scale sample set, thereby generating a large-scale branch model. The method is used to train the small-scale branch in a dual-scale branch neural network model using a small-scale sample set to generate a small-scale branch model; and to generate a final prediction value, which includes the prediction value of the large-scale branch model for the large-scale sample set and the prediction value of the small-scale branch model for the small-scale sample set. The security review module is used to conduct security reviews of real-time time synchronization information based on the final predicted value.
8. The time-synchronized information security review device based on dual-scale feature collaboration according to claim 7, characterized in that, Also includes: The local clock update module is used to update the local clock of the time synchronization system based on the results of the security review.
9. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions executable by the at least one processor, which executes the instructions stored in the memory to perform the method as described in any one of claims 1-6.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store instructions that, when executed, cause the method as described in any one of claims 1-6 to be implemented.
