CN113094860B - Industrial control network flow modeling method based on attention mechanism - Google Patents
Industrial control network flow modeling method based on attention mechanism Download PDFInfo
- Publication number
- CN113094860B CN113094860B CN202110475603.7A CN202110475603A CN113094860B CN 113094860 B CN113094860 B CN 113094860B CN 202110475603 A CN202110475603 A CN 202110475603A CN 113094860 B CN113094860 B CN 113094860B
- Authority
- CN
- China
- Prior art keywords
- industrial control
- control network
- flow
- network flow
- attention mechanism
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/18—Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/02—CAD in a network environment, e.g. collaborative CAD or distributed simulation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Abstract
The invention discloses an industrial control network flow modeling method based on an attention mechanism, which is used for collecting communication flow between an upper computer and a lower computer in an industrial control network to generate an industrial control network flow matrix, constructing a neural network frame to model the industrial control network flow, extracting industrial control network flow characteristics, and predicting the flow at the next moment of the industrial control network by inputting historical industrial control network flow. Comprising the following steps: the method comprises the steps of collecting and preprocessing flow, collecting and preprocessing industrial control network flow data, constructing a model, extracting spatial characteristics of industrial control network flow through a convolutional neural network, extracting industrial control network flow time characteristics through a cyclic neural network, extracting important characteristics of industrial control network flow through an attention mechanism, and completing flow prediction of the industrial control network through input historical flow. According to the invention, by constructing the industrial control network flow modeling method based on the attention mechanism, the attention mechanism is introduced to further extract important characteristics contained in the industrial control network flow, and a new thought is provided for modeling the industrial control network flow.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to an industrial control network flow modeling method based on an attention mechanism.
Background
In recent years, with the continuous development and progress of computer and internet of things technology and the wide application of intelligent control equipment in industry, the industrial control network structure is more complex, the industrial control system is more flexible and variable, because the industrial control network needs a more stable operation environment, equipment nodes of the industrial control network are not changed frequently, but the real-time requirement is very high, analysis and modeling of the industrial control network flow system are particularly important, so that future trend of flow is predicted through historical flow of the industrial control network, on one hand, the industrial control network flow modeling can be used for constructing flow similar to real attack flow or scene flow business, data can be provided for industrial control network security test or network security experiment, on the other hand, the attack to the industrial control network is more frequent, and national infrastructure construction is jeopardized, and on the other hand, the industrial control network flow modeling can be used for providing reference for analysis of abnormal flow of the industrial control network.
The industrial control network flow has the characteristics of self-similarity, stability, burstiness and the like, the traditional industrial control network modeling method mainly uses the characteristics of time sequences to model, the main modeling method comprises an autoregressive moving average model (ARMA), a differential integration moving average autoregressive model (ARIMA), a fractional differential autoregressive summation moving average model (FARIMA), a Markov model and a derivative model thereof and the like, for example, in Haidong, liu Jiayong and the like, the industrial control network flow between an upper computer and a lower computer is analyzed and modeled through the differential integration moving average autoregressive model (ARIMA), the model has excellent performance on linear time sequences, but with the continuous development of the prior network technology, the distribution of the industrial control network flow tends to be complicated, and the traditional modeling method cannot cope with the burstiness, so that the nonlinear modeling of the industrial control network flow aiming at the industrial control network characteristics becomes a great challenge.
In recent years, neural networks have been developed rapidly and gradually applied to various fields, the industrial control network traffic is modeled by utilizing the strong processing capacity of a neural network model to solve the nonlinear problem, zhang Yansheng, li Xiwang and the like use convolutional neural networks to detect abnormal traffic of the industrial control network, common neural networks used for industrial control network traffic modeling mainly comprise convolutional neural networks (Convolutional Neural Networks CNN), automatic encoders (Stacked Autoencoder SAE), cyclic neural networks (Recursive Neural Network RNN), boltzmann machine (Restricted Bolzmann Machine RBM) and the like, but the industrial control network modeling is only carried out by adopting a single neural network, and various characteristics cannot be extracted.
Disclosure of Invention
The invention provides an industrial control network flow modeling method based on an attention mechanism, which is characterized in that the spatial characteristics of industrial control network flow are extracted through a convolutional neural network, the time characteristics of industrial control network flow are extracted through a cyclic neural network, the important characteristics of industrial control network flow are extracted through an attention mechanism, and the flow matrix between a historical industrial control network upper computer and a historical industrial control network lower computer can be finally obtained through the neural network framework.
The invention provides an industrial control network flow modeling method based on an attention mechanism, which comprises the following steps:
(1) Data acquisition and pretreatment: aiming at the communication between an upper computer and a lower computer in an industrial control network, namely the process monitoring layer comprises equipment such as an engineer station, an operator station, a data server and the like, and the communication with a field control layer equipment controller is carried out, the flow collection work is carried out, an industrial control network flow matrix M is generated, and the normalization processing is carried out.
(2) And (3) model construction: extracting spatial feature M in industrial control network flow through two-dimensional convolutional neural network s Extracting time feature M by BiLSTM t Finally, extracting important characteristics M of industrial control network flow through a multi-head attention mechanism i And modeling the industrial control network flow is completed through the full connection layer.
(3) Iteration and training: the deep learning model based on the attention mechanism is obtained by using normalized root mean square error (RNMSE) as a loss function, and by calculating gradients, back-propagating update parameters, iterating and training.
Further, the pretreatment step in the step (1) includes the following steps:
a. the communication flow between the industrial control network process monitoring layer and the field control layer is collected, the equipment required to be collected by the process monitoring layer respectively comprises an operator station, an engineer station, a monitoring server, a real-time database, an interface server and the like, and the field control layer comprises a controller for collecting the flow of all the equipment.
b. The industrial control network process monitoring layer and the field control layer equipment are respectively numbered, the process monitoring layer equipment is numbered (1, 2, …, X), the field control layer equipment is numbered (1, 2, …, Y), and the collected industrial control network flow is formed into an industrial control network flow matrix M according to a fixed sequence, wherein the representation mode is as follows, and (i, j) represents the network flow value from the process monitoring layer to the field control layer at a certain moment.
c. Setting a time interval T, continuously sampling network flow values from a process monitoring layer to a field control layer according to the time interval T, and finally generating S industrial control flow matrixes M with dimensions of X multiplied by Y if the sampling times are S.
d. And carrying out global normalization on all the acquired industrial control flow matrixes M, and unifying all network flow values to be within the range of (0, 1).
Further, the step of modeling in the step (2) includes the following steps:
a. constructing a two-dimensional convolutional neural network, inputting P X Y industrial control flow matrixes by a model, extracting features of each industrial control flow matrix by using one two-dimensional convolutional neural network, flattening the extracted two-dimensional matrixes into one-dimensional vectors, and obtaining extracted industrial control network flow space features M s 。
b. Constructing a bidirectional cyclic neural network, and inputting the bidirectional cyclic neural network into the network traffic space feature M obtained in the previous step s Extracting time features contained in the network traffic by using BiLSTM, and recombining the output of each moment into a feature matrix to obtain an extracted industrial control network traffic time feature matrix M t 。
c. B, the network flow time characteristic matrix M generated in the previous step B is generated t As input, a multi-headed self-attention mechanism is adopted, and the self-attention mechanism passes through different weight matrixes W Q 、W K 、W V Performing linear transformation to obtain a matrix Q, K, V, dividing the three matrices into H heads respectively, performing scaling click operation, and finally combining the heads to obtain an extracted industrial control network flow important feature matrix M i
d. Constructing a full connection layer, and constructing a network traffic characteristic matrix M extracted in the step C i Flattening the data into a one-dimensional vector to serve as input, constructing a layer of fully connected network, and obtaining an industrial control network flow prediction result M at the final P+1 moment P+1 Thus, the model construction is completed.
Further, the step of iterating and training in the step (3) includes the steps of:
a. input P historical industrial control flow matrixes [ M ] 1 ，M 2 ，M 3 …M P ]Outputting the attention-based industrial control network flow model based on the attention mechanism in the step 3An industrial control network traffic model of the mechanism.
b. Initializing model parameters, setting one-time trained batch T b At T b And minimizing a loss function, ensuring the stability of a network flow prediction result in order to ensure that an abnormal value does not appear in the predicted flow, optimizing a training process by using an Adam optimizer by using the loss function by adopting a normalized root mean square error (RNMSE), and performing loop iteration until the set iteration times are reached.
c. And adjusting parameters, repeating the steps, and storing a model for obtaining the optimal result as an industrial control network flow model based on the attention mechanism.
The method can model according to the historical industrial control network flow data and forecast the future industrial control flow value, and has the following advantages compared with the prior art:
1. the neural network based on the attention mechanism is used for modeling the communication network flow between the upper computer and the lower computer of the industrial control network for the first time, and the problems that the traditional method cannot cope with the burst flow and cannot accurately model the nonlinear problem are solved.
2. According to the invention, the convolutional neural network, the circulating neural network and the attention mechanism are combined to perform industrial control network flow modeling, the spatial characteristics of industrial control network flow are extracted through the convolutional neural network, the time characteristics of industrial control network flow are extracted through the circulating neural network, the important characteristics of industrial control network flow are extracted through the attention mechanism, and the flow characteristics of the industrial control network are extracted more comprehensively.
3. The invention uses the normalized root mean square error (RNMSE) as the loss function to guide the training of the neural network, thereby accelerating the training speed and improving the accuracy of the model.
Drawings
FIG. 1 is a general flow chart of the method of the present invention. Mainly comprises three parts: the system comprises an acquisition and preprocessing part, a model construction part and a model training part.
FIG. 2 is a schematic diagram of an industrial control network using the method of the present invention, showing the acquisition and preprocessing portions
Fig. 3 is a schematic diagram of an industrial control network deep learning model based on an attention mechanism, which is a model construction part of the method.
Detailed Description
The invention is described in further detail below with reference to the detailed description and the accompanying drawings.
The invention discloses an industrial control network flow modeling method based on an attention mechanism, which can predict the communication flow value between future industrial control devices through the historical network flow of the devices between an industrial control network process monitoring layer and a field control layer, and the specific flow is shown in a figure 1, and comprises the following main steps:
and 101, collecting communication flow of an industrial control network process monitoring layer and a field control layer, and carrying out pretreatment such as normalization. The method comprises the following specific steps:
(1) The schematic diagram of the industrial control network is shown in fig. 2, and the analog industrial control network comprises 3 parts, namely a process monitoring layer, a network switching layer and a field monitoring layer. The equipment that process monitoring layer needs to gather includes operator station, engineer station, control server, real-time database, interface server etc. respectively, and the network switching layer mainly contains the switch, and the site control layer includes the controller, and the flow collection process is mainly the communication flow between collection industrial control network process monitoring layer and the site control layer, gathers at the network switching layer promptly.
(2) The industrial control network process monitoring layer and the field control layer equipment are respectively numbered, the process monitoring layer equipment is numbered (1, 2, …, X), the field control layer equipment is numbered (1, 2, …, Y), and the collected industrial control network flow is formed into an industrial control network flow matrix M according to a fixed sequence, wherein the representation mode is as follows, and (i, j) represents the network flow value from the process monitoring layer to the field control layer at a certain moment.
(3) Continuously sampling network flow values from the process monitoring layer to the field control layer at time intervals of 2min, continuously sampling for a period of time, and setting the sampling times as S, and finally generating S industrial control flow matrixes M with dimensions of X multiplied by Y, and constructing an industrial control flow data set by utilizing all acquired data.
(4) Global normalization is carried out on all the acquired industrial control flow matrixes M, all network flow values are unified into a (0, 1) range, the normalization function is as follows, and an input variable x represents the industrial control network flow matrix M:
step 201, constructing a deep learning model of industrial control network flow based on an attention mechanism through a convolutional neural network, a cyclic neural network and the attention mechanism. The method comprises the following specific steps:
(5) Constructing a two-dimensional convolutional neural network, inputting 20X multiplied by Y industrial control flow matrixes into a model, extracting features of each industrial control flow matrix by using one two-dimensional convolutional neural network, flattening the extracted two-dimensional matrixes into one-dimensional vectors, and obtaining extracted industrial control network flow space features M s 。
(6) Constructing a bidirectional cyclic neural network, and inputting the bidirectional cyclic neural network into the network traffic space feature M obtained in the previous step s In order to capture the long-distance bidirectional dependency relationship, the BiLSTM is adopted to extract the time characteristics contained in the network traffic, the output at each moment is recombined into a characteristic matrix, and the extracted industrial control network traffic time characteristic matrix M is obtained t 。
(7) Extracting important characteristics of industrial control network traffic by using an attention mechanism, and generating a network traffic time characteristic matrix M in the last step B t As input, a multi-headed self-attention mechanism is adopted, and the self-attention mechanism passes through different weight matrixes W Q 、W K 、W V Performing linear transformation to obtain a matrix Q, K, V, dividing the three matrices into 3 heads respectively, performing scaling click operation, and finally combining the heads to obtain an extracted industrial control network flow important feature matrix M i
(8) Constructing a full connection layer, and performingNetwork traffic feature matrix M extracted in step C i Flattening the data into a one-dimensional vector to serve as input, constructing a layer of fully connected network, and obtaining an industrial control network flow prediction result M at the next moment 21 Thus, the model construction is completed.
Step 301, using the normalized root mean square error as a loss function for model iterative update, performing continuous iteration and training to obtain a final model, which specifically comprises the following steps:
(9) 20 historical industrial control flow matrixes [ M ] are input 1 ，M 2 ，M 3 …M 20 ]And outputting the industrial control network flow model based on the attention mechanism through the industrial control network flow model based on the attention mechanism in the step 3.
(10) Initializing model parameters, and setting one-time trained batch T b At T b In order to ensure that abnormal values do not appear in the predicted flow and ensure the stability of the network flow prediction result, the loss function adopts a normalized root mean square error (RNMSE) with the following formula: wherein the method comprises the steps ofRepresenting the predicted value of the network traffic, y representing the true value,/->Representing the mean.
The training process is optimized by using an Adam optimizer, and the iteration is circulated until the set iteration times are reached.
(11) And adjusting parameters, repeating the steps, and storing a model for obtaining the optimal result as an industrial control network flow model based on the attention mechanism.
The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the scope of the present invention are intended to be included in the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.
Claims (1)
1. An industrial control network flow modeling method based on an attention mechanism is characterized by comprising the following steps:
A. data acquisition and preprocessing, including the steps of:
a1, collecting communication traffic between a process monitoring layer and a field control layer in an industrial control network, numbering process monitoring layer equipment (1, 2, …, X), numbering field control layer equipment (1, 2, …, Y), and forming an industrial control network traffic matrix M by the collected industrial control network traffic according to a fixed sequence, wherein the expression mode is as follows, and (X, Y) represents industrial control network traffic values from the process monitoring layer to the field control layer at a certain moment:
a2, setting a time interval T, continuously sampling industrial control network flow values from a process monitoring layer to a field control layer according to the time interval T, and finally generating S industrial control flow matrixes M with dimensions of X multiplied by Y if the sampling times are S;
a3, carrying out global normalization on all the acquired industrial control flow matrixes M, unifying all industrial control network flow values into a (0, 1) range, and carrying out normalization functions as follows:
wherein the input variable x represents an industrial control network flow matrix M;
B. the model construction comprises the following steps:
b1, constructing a two-dimensional convolutional neural network, inputting P X Y industrial control flow matrixes into a model, and performing special processing on each industrial control flow matrix by using one two-dimensional convolutional neural networkExtracting features, namely flattening the extracted two-dimensional matrix into a one-dimensional vector to obtain extracted industrial control network flow space features M s ；
B2, constructing a bidirectional circulating neural network, and inputting the bidirectional circulating neural network into the industrial control network flow space characteristic M obtained in the previous step B1 s Extracting time features contained in industrial control network traffic by using BiLSTM, and recombining the output of each moment into a feature matrix to obtain an extracted industrial control network traffic time feature matrix M t ；
B3, extracting important characteristics of industrial control network traffic by using an attention mechanism, and generating a network traffic time characteristic matrix M in the last step t As input, a multi-headed self-attention mechanism is adopted, and the self-attention mechanism passes through different weight matrixes W Q 、W K 、W V Performing linear transformation to obtain a matrix Q, K, V, dividing the three matrices into H heads respectively, performing scaling click operation, and finally combining the heads to obtain an extracted industrial control network flow important feature matrix M i ；
B4, constructing a full connection layer, and constructing the industrial control network flow characteristic matrix M extracted in the step B3 i Flattening the data into a one-dimensional vector to serve as input, constructing a layer of fully connected network, and obtaining an industrial control network flow prediction result M at the final P+1 moment P+1 Thus, the model construction is completed;
C. iteration and training, comprising the steps of:
c1, P historical industrial control flow matrixes [ M ] are input 1 ,M 2 ,M 3 …M P ]Outputting the industrial control network flow model based on the attention mechanism through the industrial control network flow model based on the attention mechanism in the step B;
c2, initializing model parameters, and setting one-time trained batch T b At T b In order to ensure that abnormal values do not appear in the predicted flow, the stability of the industrial control network flow prediction result is ensured, and the loss function adopts normalized root mean square error, and the formula is as follows:
wherein the method comprises the steps ofPredicted value representing industrial control network flow, y representing true value, < >>Representing the mean value, optimizing the training process by using an Adam optimizer, and performing loop iteration until the set iteration times are reached;
and C3, adjusting parameters, repeating the steps, and storing a model for obtaining an optimal result to serve as an industrial control network flow model based on an attention mechanism.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110475603.7A CN113094860B (en) | 2021-04-29 | 2021-04-29 | Industrial control network flow modeling method based on attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110475603.7A CN113094860B (en) | 2021-04-29 | 2021-04-29 | Industrial control network flow modeling method based on attention mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113094860A CN113094860A (en) | 2021-07-09 |
| CN113094860B true CN113094860B (en) | 2023-09-01 |
Family
ID=76680728
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110475603.7A Active CN113094860B (en) | 2021-04-29 | 2021-04-29 | Industrial control network flow modeling method based on attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113094860B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113572752B (en) * | 2021-07-20 | 2023-11-07 | 上海明略人工智能(集团)有限公司 | Abnormal flow detection method and device, electronic equipment and storage medium |
| CN113691542B (en) * | 2021-08-25 | 2023-04-28 | 中南林业科技大学 | Web attack detection method and related equipment based on HTTP request text |
| CN113807027B (en) * | 2021-10-09 | 2023-08-18 | 华北电力大学（保定） | Wind turbine generator system health state evaluation model, method and system |
| CN114553718B (en) * | 2022-02-20 | 2023-04-18 | 武汉大学 | Network traffic matrix prediction method based on self-attention mechanism |
| CN115442253B (en) * | 2022-09-15 | 2023-07-18 | 西安电子科技大学 | Network flow prediction method utilizing attention mechanism |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111030889A (en) * | 2019-12-24 | 2020-04-17 | 国网河北省电力有限公司信息通信分公司 | Network traffic prediction method based on GRU model |
| CN111209933A (en) * | 2019-12-25 | 2020-05-29 | 国网冀北电力有限公司信息通信分公司 | Network traffic classification method and device based on neural network and attention mechanism |
| CN111818052A (en) * | 2020-07-09 | 2020-10-23 | 国网山西省电力公司信息通信分公司 | CNN-LSTM-based industrial control protocol homologous attack detection method |
| CN111970163A (en) * | 2020-06-30 | 2020-11-20 | 网络通信与安全紫金山实验室 | Network flow prediction method of LSTM model based on attention mechanism |
| CN112350876A (en) * | 2021-01-11 | 2021-02-09 | 南京信息工程大学 | Network flow prediction method based on graph neural network |
| CN112418361A (en) * | 2021-01-22 | 2021-02-26 | 杭州木链物联网科技有限公司 | Industrial control system anomaly detection method and device based on deep learning |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10855706B2 (en) * | 2016-10-11 | 2020-12-01 | Battelle Memorial Institute | System and methods for automated detection, reasoning and recommendations for resilient cyber systems |
| US20200204571A1 (en) * | 2018-12-19 | 2020-06-25 | AVAST Software s.r.o. | Malware detection in network traffic time series |
-
2021
- 2021-04-29 CN CN202110475603.7A patent/CN113094860B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111030889A (en) * | 2019-12-24 | 2020-04-17 | 国网河北省电力有限公司信息通信分公司 | Network traffic prediction method based on GRU model |
| CN111209933A (en) * | 2019-12-25 | 2020-05-29 | 国网冀北电力有限公司信息通信分公司 | Network traffic classification method and device based on neural network and attention mechanism |
| CN111970163A (en) * | 2020-06-30 | 2020-11-20 | 网络通信与安全紫金山实验室 | Network flow prediction method of LSTM model based on attention mechanism |
| CN111818052A (en) * | 2020-07-09 | 2020-10-23 | 国网山西省电力公司信息通信分公司 | CNN-LSTM-based industrial control protocol homologous attack detection method |
| CN112350876A (en) * | 2021-01-11 | 2021-02-09 | 南京信息工程大学 | Network flow prediction method based on graph neural network |
| CN112418361A (en) * | 2021-01-22 | 2021-02-26 | 杭州木链物联网科技有限公司 | Industrial control system anomaly detection method and device based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 基于烟草工业网络流量监测分析与安全防护应用;孟瑾;;网络安全技术与应用(第02期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113094860A (en) | 2021-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113094860B (en) | Industrial control network flow modeling method based on attention mechanism | |
| CN108448610B (en) | Short-term wind power prediction method based on deep learning | |
| CN113313947B (en) | Road condition evaluation method of short-term traffic prediction graph convolution network | |
| CN110163429B (en) | Short-term load prediction method based on similarity day optimization screening | |
| CN110377942B (en) | Multi-model space-time modeling method based on finite Gaussian mixture model | |
| CN109523155B (en) | Power grid risk assessment method of Monte Carlo and least square support vector machine | |
| CN111127246A (en) | Intelligent prediction method for transmission line engineering cost | |
| CN105893669A (en) | Global simulation performance predication method based on data digging | |
| CN112257911A (en) | TCN multivariate time sequence prediction method based on parallel space-time attention mechanism | |
| CN108182316B (en) | Electromagnetic simulation method based on artificial intelligence and electromagnetic brain thereof | |
| CN112926256B (en) | Soil heavy metal content prediction method and system based on deep reinforcement learning | |
| Zhou et al. | Network traffic prediction method based on improved echo state network | |
| CN113988449A (en) | Wind power prediction method based on Transformer model | |
| CN112381279B (en) | Wind power prediction method based on VMD and BLS combined model | |
| Li et al. | A deep learning framework for non-stationary time series prediction | |
| CN113128666A (en) | Mo-S-LSTMs model-based time series multi-step prediction method | |
| CN116599857A (en) | Digital twin application system suitable for multiple scenes of Internet of things | |
| CN112149896A (en) | Attention mechanism-based mechanical equipment multi-working-condition fault prediction method | |
| CN116663419A (en) | Sensorless equipment fault prediction method based on optimized Elman neural network | |
| CN113835964B (en) | Cloud data center server energy consumption prediction method based on small sample learning | |
| CN115907154A (en) | Industrial internet prediction method and system based on frequency domain and long-term and short-term feature fusion | |
| CN113095596A (en) | Photovoltaic power prediction method based on multi-stage Gate-SA-TCN | |
| CN113112075B (en) | Internal combustion engine noise prediction method based on VMD and NARX | |
| CN113762351B (en) | Air quality prediction method based on deep transition network | |
| CN111950691A (en) | Reinforced learning strategy learning method based on potential action representation space |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |