CN115964258A - Internet of things network card abnormal behavior grading monitoring method and system based on multi-time sequence analysis - Google Patents
Internet of things network card abnormal behavior grading monitoring method and system based on multi-time sequence analysis Download PDFInfo
- Publication number
- CN115964258A CN115964258A CN202211721988.1A CN202211721988A CN115964258A CN 115964258 A CN115964258 A CN 115964258A CN 202211721988 A CN202211721988 A CN 202211721988A CN 115964258 A CN115964258 A CN 115964258A
- Authority
- CN
- China
- Prior art keywords
- data
- training
- internet
- trained
- monitoring
- 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.)
- Pending
Links
Images
Classifications
-
- 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
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a multi-time sequence analysis-based method and a multi-time sequence analysis-based system for monitoring abnormal behaviors of an Internet of things card in a grading manner. The hierarchical monitoring method for the abnormal behavior of the Internet of things card based on the multi-time sequence analysis comprises the following steps: long-distance monitoring is carried out on the user behavior data of the Internet of things card in the complete life cycle by using a gated circulation unit model, suspected abnormal behavior data are intercepted and classified; analyzing the content before and after the suspected abnormal behavior by using a bidirectional circulation neural network model, realizing classification and weight distribution of events, and acquiring quality risk monitoring data; and auditing the competitive product risk monitoring data by using a Transformer model, and outputting a competitive product data result. The invention utilizes a plurality of models to detect the user behavior data of the Internet of things card, enhances the monitoring period, can complete long-distance monitoring, and simultaneously improves the discovery rate and the accuracy rate of abnormal events.
Description
Technical Field
The invention relates to the technical field of Internet of things, in particular to a multi-time sequence analysis-based method and a multi-time sequence analysis-based system for monitoring abnormal behaviors of an Internet of things.
Background
The Internet of things card is a mobile communication access service provided by an operator for Internet of things users based on an Internet of things public service network. For the usage data of users of 31 provinces of Internet of things in China, the amount of data generated each day is huge. At present, the use data of the internet of things card user realizes the discovery of abnormal behaviors through the performance of a stack server, the algorithm model is single and is easy to bypass behaviors, and the discovery rate and the accuracy rate of the abnormal behaviors are low.
Disclosure of Invention
The invention aims to provide a method and a system for monitoring abnormal behaviors of an Internet of things card based on multi-time sequence analysis, which can improve the discovery rate and the accuracy rate of the abnormal behaviors.
In order to solve the technical problems, the invention adopts the following technical scheme:
a hierarchical monitoring method for abnormal behaviors of an Internet of things card based on multi-time sequence analysis comprises the following steps: long-distance monitoring is carried out on the user behavior data of the Internet of things card in the complete life cycle by utilizing a trained gate control cycle unit model, suspected abnormal behavior data are intercepted and classified; aiming at suspected abnormal behavior data, analyzing the contents before and after the behavior by using a trained bidirectional circulation neural network model, realizing classification and weight distribution of events, and acquiring fine product risk monitoring data; and (4) auditing the competitive product risk monitoring data by using the trained Transformer model, and outputting a competitive product data result.
A multi-time sequence analysis-based Internet of things network card abnormal behavior grading monitoring system comprises a suspected abnormal behavior identification module, a suspected abnormal behavior analysis module and an abnormal behavior auditing module; the suspected abnormal behavior recognition module is used for utilizing the trained gating cycle unit model to carry out long-distance monitoring on the user behavior data of the Internet of things card in the complete life cycle, intercepting the suspected abnormal behavior data and carrying out classification; the suspected abnormal behavior analysis module is used for analyzing the contents before and after the behavior by utilizing a trained bidirectional circulation neural network model aiming at the suspected abnormal behavior data, realizing the classification and weight distribution of events and acquiring quality risk monitoring data; and the abnormal behavior auditing module is used for auditing the competitive product risk monitoring data by using the trained Transformer model and outputting a competitive product data result.
The invention has the beneficial technical effects that: according to the method and the system for monitoring the abnormal behaviors of the Internet of things card based on the multi-time-sequence analysis in a grading manner, the abnormal behavior data of the Internet of things card in a long-distance time sequence is found through the gated circulation unit model, the contents of the front and back actions of the intercepted suspected abnormal behavior data are analyzed and graded and the weight is distributed through the bidirectional circulation neural network model, and finally the middle and high risk events in the competitive risk monitoring data are accurately positioned through the Transformer model, so that the output of the result is completed. The invention utilizes a plurality of time sequence analysis models to detect the user behavior data of the Internet of things card in stages, enhances the monitoring period, can complete long-distance monitoring, and simultaneously improves the discovery rate and the accuracy rate of abnormal events.
Drawings
FIG. 1 is a schematic flow chart of a multi-timing analysis-based hierarchical monitoring method for abnormal behavior of an Internet of things card according to the present invention;
FIG. 2 is a schematic diagram of a gated loop cell model according to the present invention;
FIG. 3 is a schematic diagram of a bidirectional recurrent neural network model of the present invention;
FIG. 4 is a schematic structural diagram of a Transformer model according to the present invention;
FIG. 5 is a schematic structural diagram of an encoder and a decoder of the Transformer model according to the present invention;
FIG. 6 is a schematic diagram of a self-attention matrix operation method according to the present invention;
fig. 7 is a schematic structural diagram of a hierarchical monitoring system for abnormal behavior of an internet of things card based on multi-time-sequence analysis according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more clearly understood by those skilled in the art, the present invention is further described with reference to the accompanying drawings and examples.
As shown in fig. 1, in an embodiment of the present invention, a hierarchical monitoring method for abnormal behavior of an internet of things card based on multiple time series analysis includes the following steps:
s10, performing long-distance monitoring on the user behavior data of the Internet of things card in the complete life cycle by using the trained gate control cycle unit model, intercepting suspected abnormal behavior data and grading;
s20, aiming at the suspected abnormal behavior data, analyzing the contents before and after the behavior by using a trained bidirectional circulation neural network model, realizing classification and weight distribution of events, and acquiring fine product risk monitoring data;
and S30, auditing the competitive product risk monitoring data by using the trained Transformer model, and outputting a competitive product data result.
In step S10:
the gated cyclic unit model takes sequence data as input, recursion is carried out in the evolution direction of the sequence, all nodes are connected in a chain manner to form a cyclic neural network, the unit state is increased by combining the characteristic of long risk event period of the Internet of things, and a long-and-short-term memory network is formed to meet the requirement of long-distance dependence of data.
The gating cycle unit model is obtained by pre-training based on training data, and the training process is as follows:
s11, collecting normal behavior data and alarm data of the multiple provincial desensitized Internet of things card as training data, and dividing the training data into a training set and a test set.
Collecting normal behavior data and alarm data of the IOT card after a plurality of provincial desensitizations, and generating training data after carrying out screening, duplication removal, denoising and feature extraction on the collected normal behavior data and alarm data of the IOT card. After the training data is generated, randomly segmenting the training data into a training set and a testing set according to the proportion of 8. The behavior data of the Internet of things card comprises information such as voice phone list quantity, short message phone list quantity, flow phone list quantity, whether a machine card is separated or not, whether the machine card is used across regions or not, whether the machine card is used in a roaming sensitive area or not and the like, and information such as normal use, unreasonable use, abnormal flow use, abnormal short message use and the like of a label corresponding to the data.
And S12, sending the training set into a gate control cycle unit model to be trained for training, and generating the trained gate control cycle unit model through test set verification.
FIG. 2 shows a schematic structural diagram of a gated cyclic unit model, and as shown in FIG. 2, the present invention modifies the calculation of hidden states in a cyclic neural network. The inputs of the reset gate and the update gate in the gated cycle cell model are both current time step input X t Hidden state H with last time step t-1 And the output is obtained by calculating the full connection layer with the activation function being the sigmoid function.
Setting the number of hidden units as h, and setting the small batch input X of time step t t ∈R n*d (n samples, d inputs) and hidden state H of last time step t-1 ∈R n*h . Reset gate R t ∈R n*h And a refresh door Z t ∈R n*h Is calculated as follows:
R t ＝σ(X t W xr +H t-1 W hr +b r )
Z t ＝σ(X t W xz +H t-1 W hz +b z )
the sigmoid function can transform the value of an element between 0 and 1, thereby capturing the dependence relationship with larger time step distance in the time sequence and controlling the information flow through a learned gate.
The gated circulation unit model adopts fewer parameters, so that the risk of overfitting can be effectively reduced.
In step S20:
the bidirectional circulation neural network model is obtained by pre-training based on training data, and the training process is as follows:
s21, collecting normal behavior data and alarm data of the multiple provincial desensitized Internet of things card as training data, and dividing the training data into a training set and a test set.
Collecting normal behavior data and alarm data of the multiple provincial desensitized internet access cards, and performing screening, duplicate removal, denoising and feature extraction on the collected normal behavior data and alarm data of the internet access cards to generate training data. After the training data are generated, randomly dividing the training data into a training set and a test set according to the proportion of 8. The behavior data of the Internet of things card comprises information such as voice phone list quantity, short message phone list quantity, flow phone list quantity, whether the phone and the card are separated, whether the phone and the card are used across regions, whether the phone and the card are used in a roaming sensitive region and the like, and information such as normal use, unreasonable use, abnormal flow use, abnormal short message use and the like of a label corresponding to the data.
And S22, sending the training set into a to-be-trained bidirectional circulation neural network model for training, and finally generating a trained bidirectional circulation neural network model through verification of the test set.
Fig. 3 shows a structural diagram of a bidirectional recurrent neural network model, as shown in fig. 3, the bidirectional recurrent neural network connects two opposite hidden layers to the same output, and the output layer can receive information from the forward direction and the backward direction simultaneously based on generative deep learning. Through this structure, complete past and future context data information is provided to each point in the input sequence at the output layer. And (3) repeatedly utilizing six unique weights in each time step, wherein the six weights respectively correspond to: input to the forward and backward hidden layers (w 1, w 3), hidden layer to hidden layer itself (w 2, w 5), forward and backward hidden layer to output layer (w 4, w 6), finally forming a bidirectional recurrent neural network which evolves along time.
Forward estimation:
for the hidden layers of the bidirectional recurrent neural network, forward estimation is the same as that of the unidirectional recurrent neural network, except that the input sequences are in opposite directions for the two hidden layers, and the output layer is not updated until the two hidden layers process all the input sequences:
for t＝1 to T do
Forward pass for the forward hidden layer,storing activations at each timestep for t＝Tto 1do
Forward pass for the backward hidden layer,storing activations at each timestep for all t,in any order do
Forward pass for the output layer,using the stored activations from bothhidden layers.
backward estimation:
the back-calculation of a bi-directional recurrent neural network is similar to the back-propagation of a standard recurrent neural network through time, except that all output layer delta terms are first calculated and then returned to the hidden layers in two different directions:
for all t,in any order do
Backward pass forthe output layer,storing terms at each timestep for t＝Tto 1do
BPTT backward pass for the forward hidden layer,using the stored 8terms from the output layer for t＝1to Tdo
BPTT backward pass for the backward hidden layer,using the stored 8terms from the output layer.
the bidirectional circulation neural network model carries out bidirectional circulation calculation on the complete sequence by taking time as an axis, carries out forward and backward behavior analysis on sequence behavior data of the Internet of things card, analyzes complete events and reasoning behavior, and can discover a complex scene model.
In step S30:
the Transformer model is obtained by pre-training based on training data, and the training process is as follows:
s31, collecting normal behavior data and alarm data of the IOT card after the provincial desensitization as training data, and dividing the training data into a training set and a test set.
Collecting normal behavior data and alarm data of the IOT card after a plurality of provincial desensitizations, and generating training data after carrying out screening, duplication removal, denoising and feature extraction on the collected normal behavior data and alarm data of the IOT card. After the training data is generated, randomly segmenting the training data into a training set and a testing set according to the proportion of 8. The behavior data of the Internet of things card comprises information such as voice phone list quantity, short message phone list quantity, flow phone list quantity, whether a machine card is separated or not, whether the machine card is used across regions or not, whether the machine card is used in a roaming sensitive area or not and the like, and information such as normal use, unreasonable use, abnormal flow use, abnormal short message use and the like of a label corresponding to the data.
And S32, sending the training set into a Transformer model to be trained for training, and generating the trained Transformer model through verification of the test set.
Fig. 4 shows a schematic structural diagram of a Transformer model, and as shown in fig. 4, a black box made by the Transformer model can realize translation of an input language into another language, and the black box is composed of 2 parts, namely an encoding component and a decoding component, which are respectively composed of N encoders and decoders, and the accuracy of the model is improved by calculation of multi-encoding and multi-decoding.
As shown in fig. 5, the encoder includes a self-attention layer and a feed-forward neural network. As shown in fig. 6, the process of the self-attention matrix operation method is as follows: 1) Inputs (Thinking Machines); 2) Encoding each action (note: except the 0 th encoder, other encoders do not need word embedding, and it can directly take the output of the previous encoder as input (matrix R)); 3) Dividing the self-attention into 8 heads, and multiplying the matrix X or the matrix R by each weight matrix; 4) Calculating attention through the output query/key/value (Q/K/V) matrix; 5) All the heads of attention are stitched together and multiplied by a weight matrix W.
The internet of things card behavior data passes through the self-attention layer of the encoder, which helps the encoder focus on other behaviors of the input data as each behavior is encoded. The output from the attention layer is passed into a feedforward neural network, and a window is a one-dimensional convolutional neural network of behavior.
Through the Transformer model, each element can interact with global information like CNN (convolutional neural network), the global distance is ignored, meanwhile, a model with better interpretability can be generated due to self attention, attention distribution is checked from the model, each attention head can learn to execute different tasks, and the problem that the RNN (convolutional neural network) model cannot perform parallel calculation is solved.
The complete abnormal behavior event comprises a complete life cycle, similar preamplifiers may exist in the same event type, and finally, the complete abnormal event is formed by combining one or more preamplifiers with the abnormal behavior. In the grading monitoring process of the abnormal behaviors of the Internet of things card, for continuously receiving the user behavior data of the Internet of things card, analyzing through a gate control cycle unit model, modifying a hidden state calculation mode in a cycle neural network by building a reset gate and an update gate, discovering the long-distance risk behaviors in time, carrying out preliminary risk grading and deepening management and control according to the types of the risk behaviors, and intercepting suspected abnormal behavior data according to the preliminary screening result; calculating forward and backward through a bidirectional cyclic neural network model aiming at intercepted suspected abnormal behavior data, reasoning forward and backward events of event nodes, and grading the threat degree of the events; and after grading is completed, aiming at the medium and high-risk data, a black box is made through a Transformer model, the user behavior data of the Internet of things network card is used as input, self-attention matrix operation is performed through an encoder, concurrent multidimensional calculation is performed by taking the global situation as a range, and accurate grading of the user risk event is completed.
As shown in fig. 7, the present invention further provides a hierarchical monitoring system for abnormal behaviors of an internet of things card based on multi-time-series analysis, which includes a suspected abnormal behavior identification module 10, a suspected abnormal behavior analysis module 20, and an abnormal behavior audit module 30.
The suspected abnormal behavior recognition module 10 is configured to perform long-distance monitoring on the user behavior data of the internet of things card in the complete life cycle by using the trained gated cycle unit model, intercept the suspected abnormal behavior data, and perform classification, that is, the suspected abnormal behavior recognition module 10 is configured to execute step S10 in the hierarchical monitoring method for the abnormal behavior of the internet of things card based on the multi-time-sequence analysis in the embodiment shown in fig. 1.
The suspected abnormal behavior analysis module 20 is configured to analyze the contents before and after the behavior by using the trained bidirectional cyclic neural network model for the suspected abnormal behavior data, so as to implement classification and weight distribution of the event, and obtain top-quality risk monitoring data, that is, the suspected abnormal behavior analysis module 20 is configured to execute step S20 in the hierarchical monitoring method for the abnormal behavior of the internet of things card based on the multi-time-sequence analysis in the embodiment shown in fig. 1.
The abnormal behavior auditing module 30 is configured to audit the competitive product risk monitoring data by using the trained Transformer model, and output a competitive product data result, that is, the abnormal behavior auditing module 30 is configured to execute step S30 in the hierarchical monitoring method for the abnormal behavior of the internet of things card based on the multi-time-sequence analysis in the embodiment shown in fig. 1.
According to the hierarchical monitoring method and system for the abnormal behaviors of the Internet of things card based on multi-time-sequence analysis, the abnormal behavior data of the Internet of things card in a long-distance time sequence is found through a gated circulation unit model, the contents of the front and back actions of the intercepted suspected abnormal behavior data are analyzed and graded and the weight is distributed through a bidirectional circulation neural network model, and finally the middle and high risk events in the fine risk monitoring data are accurately positioned through a Transformer model, so that the output of the result is completed. The user behavior data of the Internet of things card is detected by utilizing the plurality of time sequence analysis models, the monitoring period is prolonged, long-distance monitoring can be completed, the discovery rate and the accuracy rate of abnormal events are improved, in addition, the event weight can be regulated and controlled through the classification of the events, and therefore the resource utilization rate is improved.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Various equivalent changes and modifications can be made by those skilled in the art based on the above embodiments, and all equivalent changes and modifications within the scope of the claims should fall within the protection scope of the present invention.
Claims (6)
1. A hierarchical monitoring method for abnormal behaviors of an Internet of things card based on multi-time sequence analysis is characterized in that the hierarchical monitoring method for the abnormal behaviors of the Internet of things card based on the multi-time sequence analysis comprises the following steps:
monitoring the user behavior data of the Internet of things card in a long distance in a complete life cycle by using a trained gating cycle unit model, intercepting suspected abnormal behavior data and grading;
aiming at suspected abnormal behavior data, analyzing the contents before and after the behavior by using a trained bidirectional circulation neural network model, realizing classification and weight distribution of events, and acquiring fine product risk monitoring data;
and auditing the competitive product risk monitoring data by using the trained Transformer model, and outputting a competitive product data result.
2. The Internet of things card abnormal behavior grading monitoring method based on multi-time sequence analysis as claimed in claim 1, wherein the training method of the gating cycle unit model comprises the following steps:
collecting normal behavior data and alarm data of the IOT card after a plurality of provincial desensitizations as training data, and dividing the training data into a training set and a test set;
and sending the training set into a gate control cycle unit model to be trained for training, and finally generating the trained gate control cycle unit model after verification of the test set.
3. The Internet of things card abnormal behavior grading monitoring method based on multi-time sequence analysis as claimed in claim 1, wherein the training method of the gating cycle unit model comprises the following steps:
collecting normal behavior data and alarm data of the IOT card after a plurality of provincial desensitizations;
screening, de-duplication, de-noising and feature extraction are carried out on the collected normal behavior data and alarm data of the Internet of things card to be used as training data;
randomly segmenting training data into a training set and a test set, wherein the proportion of the training set to the test set is 8;
and sending the training set into a gate control cycle unit model to be trained for training, and generating the trained gate control cycle unit model through test set verification.
4. The Internet of things card abnormal behavior grading monitoring method based on multi-time sequence analysis as claimed in claim 1, wherein the training method of the bidirectional cyclic neural network model comprises the following steps:
collecting normal behavior data and alarm data of the IOT card subjected to province desensitization as training data, and dividing the training data into a training set and a test set;
and sending the training set into a bidirectional cyclic neural network model to be trained for training, and finally generating the trained bidirectional cyclic neural network model after verification of the test set.
5. The hierarchical monitoring method for the abnormal behavior of the internet of things card based on the multi-time-sequence analysis as claimed in claim 1, wherein the training method of the transform model comprises the following steps:
collecting normal behavior data and alarm data of the IOT card after a plurality of provincial desensitizations as training data, and dividing the training data into a training set and a test set;
and sending the training set into a Transformer model to be trained for training, and generating the trained Transformer model through verification of the test set.
6. The utility model provides a hierarchical monitoring system of thing networking card unusual behavior based on analysis of many chronologies, its characterized in that, the hierarchical monitoring system of thing networking card unusual behavior based on analysis of many chronologies includes:
the suspected abnormal behavior recognition module is used for utilizing the trained gate control cycle unit model to carry out long-distance monitoring on the user behavior data of the Internet of things card in the complete life cycle, intercepting the suspected abnormal behavior data and grading the data;
the suspected abnormal behavior analysis module is used for analyzing the contents before and after the behavior by utilizing a trained bidirectional circulation neural network model aiming at the suspected abnormal behavior data, realizing the classification and weight distribution of events and acquiring the fine product risk monitoring data;
and the abnormal behavior auditing module is used for auditing the competitive product risk monitoring data by using the trained Transformer model and outputting a competitive product data result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211721988.1A CN115964258A (en) | 2022-12-30 | 2022-12-30 | Internet of things network card abnormal behavior grading monitoring method and system based on multi-time sequence analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211721988.1A CN115964258A (en) | 2022-12-30 | 2022-12-30 | Internet of things network card abnormal behavior grading monitoring method and system based on multi-time sequence analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115964258A true CN115964258A (en) | 2023-04-14 |
Family
ID=87358457
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211721988.1A Pending CN115964258A (en) | 2022-12-30 | 2022-12-30 | Internet of things network card abnormal behavior grading monitoring method and system based on multi-time sequence analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115964258A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117150407A (en) * | 2023-09-04 | 2023-12-01 | 国网上海市电力公司 | Abnormality detection method for industrial carbon emission data |
| CN117421684A (en) * | 2023-12-14 | 2024-01-19 | 易知谷科技集团有限公司 | Abnormal data monitoring and analyzing method based on data mining and neural network |
-
2022
- 2022-12-30 CN CN202211721988.1A patent/CN115964258A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117150407A (en) * | 2023-09-04 | 2023-12-01 | 国网上海市电力公司 | Abnormality detection method for industrial carbon emission data |
| CN117421684A (en) * | 2023-12-14 | 2024-01-19 | 易知谷科技集团有限公司 | Abnormal data monitoring and analyzing method based on data mining and neural network |
| CN117421684B (en) * | 2023-12-14 | 2024-03-12 | 易知谷科技集团有限公司 | Abnormal data monitoring and analyzing method based on data mining and neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110896381B (en) | Deep neural network-based traffic classification method and system and electronic equipment | |
| Du et al. | GAN-based anomaly detection for multivariate time series using polluted training set | |
| CN106528874B (en) | The CLR multi-tag data classification method of big data platform is calculated based on Spark memory | |
| CN115964258A (en) | Internet of things network card abnormal behavior grading monitoring method and system based on multi-time sequence analysis | |
| CN113157800B (en) | Identification method for discovering dynamic target in air in real time | |
| CN115688035A (en) | Time sequence power data anomaly detection method based on self-supervision learning | |
| CN112087442B (en) | Time sequence related network intrusion detection method based on attention mechanism | |
| Karimi-Bidhendi et al. | Scalable classification of univariate and multivariate time series | |
| CN112491891B (en) | Network attack detection method based on hybrid deep learning in Internet of things environment | |
| Dou et al. | Pc 2 a: predicting collective contextual anomalies via lstm with deep generative model | |
| Legrand et al. | Study of autoencoder neural networks for anomaly detection in connected buildings | |
| CN115587335A (en) | Training method of abnormal value detection model, abnormal value detection method and system | |
| CN115600137A (en) | Multi-source domain variable working condition mechanical fault diagnosis method for incomplete category data | |
| CN116520806A (en) | Intelligent fault diagnosis system and method for industrial system | |
| CN116340524A (en) | Method for supplementing small sample temporal knowledge graph based on relational adaptive network | |
| Sheng et al. | Network traffic anomaly detection method based on chaotic neural network | |
| CN113553918B (en) | Machine ticket issuing character recognition method based on pulse active learning | |
| Huo et al. | Traffic anomaly detection method based on improved GRU and EFMS-Kmeans clustering | |
| CN114745335B (en) | Network traffic classification device, storage medium and electronic equipment | |
| CN115643153A (en) | Alarm correlation analysis method based on graph neural network | |
| Zhou et al. | Unsupervised anomaly detection approach for multivariate time series | |
| Jin et al. | An Adaptive Anomaly Detection Method for Cloud Computing System | |
| Ling et al. | Graph Attention Mechanism-Based Method for Tracing APT Attacks in Power Systems | |
| Durbeck et al. | Kalman filter driven estimation of community structure in time varying graphs | |
| CN114124676B (en) | Fault root positioning method and system for network intelligent operation and maintenance system |
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 |