Database abnormal flow detection method and system based on improved 1DCNN network

By improving the combination of 1DCNN and BiLSTM, using residual connection and adaptive hybrid pooling, the information loss problem of traditional 1DCNN in complex nonlinear traffic mode detection is solved, and the accuracy and robustness of abnormal traffic detection is improved, and the realization of Better feature extraction and pattern recognition.

CN119030760AActive Publication Date: 2024-11-26BEIJING XINSHU TECH CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411066969.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-26
Estimated Expiration
2044-08-05

AI Technical Summary

Technical Problem

Traditional one-dimensional convolutional neural networks are prone to loss of critical information when processing complex nonlinear traffic patterns, making it difficult to provide sufficiently accurate abnormal traffic detection results. In addition, the single bidirectional long and short-term memory network has weak spatial feature extraction capabilities, making it difficult to accurately Identify complex exception patterns.

Method used

Improved one-dimensional convolutional neural network (1DCNN) by introducing residual connections and adaptive hybrid pooling, combining bidirectional long and short-term memory network (BiLSTM), adopting attention mechanism and Dropout to prevent overfitting, enhancing the model to complex nonlinearity The adaptability of the mode improves the accuracy of abnormal flow detection.

Benefits of technology

Effectively capture the spatial and temporal characteristics of the data, improve the network's expression ability and prediction accuracy, enhance the detection ability of complex nonlinear traffic patterns, and improve the accuracy and robustness of abnormal traffic detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119030760A_ABST
    Figure CN119030760A_ABST
Patent Text Reader

Abstract

The invention provides a database abnormal flow detection method and system based on an improved 1DCNN network, and the system comprises a network flow data collection module, a network flow data transmission and processing module, an abnormal flow detection module, a detection result visual display module, and a system management module. Wherein the abnormal traffic detection module adopts an improved 1DCNN algorithm to carry out detection in a nonlinear traffic mode, and carries out single traffic session detection and malicious event alarm. After the method is adopted, the characteristics of the data can be deeply captured, the problem of key information loss of the existing 1DCNN in convolution calculation is solved, the prediction accuracy of a network model is improved, cross-layer information transmission is introduced in a residual connection mode, the network is helped to better capture a complex nonlinear flow mode, and the expression ability of the network is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a database abnormal traffic detection method, system, device and readable storage medium based on an improved one-dimensional convolutional neural network and a bidirectional long short-term memory network, and belongs to the field of database detection. Background Art

[0002] With the rapid development of network technology, abnormal traffic detection plays a key role in maintaining network security and stability. Abnormal traffic often indicates the existence of potential network attacks or failures, so early detection and response to abnormal traffic is crucial to ensure the normal operation of the network. At present, a variety of methods have emerged in abnormal traffic detection technology, including different methods based on statistics, machine learning, and deep learning.

[0003] Traditional abnormal traffic detection methods have some shortcomings when dealing with complex nonlinear traffic patterns. Although the commonly used one-dimensional convolutional neural network (1DCNN) can effectively extract the spatial features of data, its convolution operation easily leads to the loss of key information, which limits the network's expressive ability. In particular, in the face of increasingly complex and hidden network attack methods, the traditional 1DCNN performs poorly and is difficult to provide sufficiently accurate abnormal traffic detection results. 1DCNN uses convolution kernels of different sizes to perform convolution operations on the input data to extract the spatial features of the data. The input data format is n×1×m, which means there are n samples, each sample is a one-dimensional sequence with a length of m. The convolution operation helps to capture the characteristics of the data in depth. In the convolution calculation of the convolution layer of 1DCNN, the weights and biases are directly applied after the convolution operation, which easily causes the loss of important information during the convolution process and the network capability is weak. Therefore, in order to improve the performance, it is necessary to improve the traditional 1DCNN and introduce a more effective information transmission mechanism to better capture complex nonlinear traffic patterns, thereby improving the network's expressive ability. This improvement is of great significance for responding to the ever-evolving threats from network attackers and improving the accuracy and robustness of abnormal traffic detection.

[0004] In addition, 1DCNN mainly focuses on extracting spatial features of data, but has limitations in capturing temporal information, such as the changing trend and periodicity of traffic. The Bidirectional Long Short-Term Memory (BiLSTM) network can well capture the time dependency and contextual information in the input data; 1DCNN network is difficult to accurately identify and classify when facing complex nonlinear traffic patterns, while BiLSTM can effectively process complex temporal patterns through memory and forgetting mechanisms. However, the BiLSTM model alone has a weak ability to extract spatial features of data. Database traffic data may contain multiple dimensions and complex spatial patterns, which are critical for anomaly detection. The lack of spatial feature extraction capabilities may lead to inaccurate recognition of certain abnormal patterns. Therefore, the combination of 1DCNN and BiLSTM can make full use of the advantages of both, enhance the model's adaptability to complex nonlinear patterns, and improve the accuracy of abnormal traffic detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 This is the system framework diagram.

[0006] Figure 2 This is the algorithm flow chart. Summary of the invention

[0007] In order to solve the above technical problems, the present invention proposes a database abnormal traffic detection method based on an improved 1DCNN network, and the specific steps are as follows:

[0008] (1) Network traffic data collection;

[0009] (2) Transmit and process network traffic data;

[0010] (3) Adopt the improved 1DCNN algorithm to perform detection under nonlinear traffic mode, and perform single traffic session detection and malicious event alarm;

[0011] (4) Visual display of test results;

[0012] (5) System management.

[0013] Furthermore, in step (3), the specific algorithm steps are as follows:

[0014] 3.1 Define the number of input categories n.

[0015] 3.2 Calculate the output of layer z Among them, ReLU is the activation function; c z-1 is the cth channel of the z-1th layer; Attention is the attention mechanism; is the input vector of the z-th layer, that is, the output vector of the i-th channel of the z-1-th layer; is the weight matrix of the convolution kernel parameters, where z represents the zth layer, i is the i-th channel of the zth layer, and c is the channel in the z-1th layer; is the bias term of the i-th channel of the z-th layer; is the output vector of the i-th channel of the z-2-th layer;

[0016] Compute the mixed pooling output of the jth neuron in the ith channel in the zth layer

[0017] Among them, σ() is the activation function, p is the width of the pooling kernel, and x z(i,t) is the input of the tth neuron in the ith channel of the zth layer;

[0018] Calculate the output of the fully connected layer z in, is the weight of the line from the ith neuron in the z-1th layer to the jth neuron in the zth layer; is the output value of the i-th neuron in the z-1 layer; is the bias term of the jth neuron in the z-1 layer; is the output value of the i-th neuron in the z-2-th layer.

[0019] Finally, the classification result is calculated, and the final category of the sample is the position of the maximum value among all values.

[0020] 3.3 Input data passes through BiLSTM twice, and Dropout is used to prevent overfitting.

[0021] 3.4 Concatenate the output data of 1DCNN and BiLSTM, and then input the concatenated data into the fully connected layer. The number of neurons in the fully connected layer is set according to the number of categories n.

[0022] 3.5 Based on the number of categories n, determine whether to use the Sigmoid loss function for binary classification or the Softmax loss function for multi-classification.

[0023] 3.6 Output a number representing the category.

[0024] 3.7 Provide classified explanations based on the output results.

[0025] Based on the above method, the present invention further proposes a database abnormal traffic detection system based on an improved 1DCNN network, the system comprising:

[0026] (1) Network traffic data collection module;

[0027] (2) Network traffic data transmission and processing module;

[0028] (3) Detection module based on improved 1DCNN algorithm, which uses improved 1DCNN algorithm to detect nonlinear traffic patterns, and performs single traffic session detection and malicious event alarm;

[0029] (4) Visual display module of test results;

[0030] (5)System management module.

[0031] Furthermore, in module (3), the improved 1DCNN algorithm steps are as follows:

[0032] 3.1 Define the number of input categories n.

[0033] 3.2 Calculate the output of layer z Among them, ReLU is the activation function; c z-1 is the cth channel of the z-1th layer; Attention is the attention mechanism; is the input vector of the z-th layer, that is, the output vector of the i-th channel of the z-1-th layer; is the weight matrix of the convolution kernel parameters, where z represents the zth layer, i is the i-th channel of the zth layer, and c is the channel in the z-1th layer; is the bias term of the i-th channel of the z-th layer; is the output vector of the i-th channel of the z-2-th layer;

[0034] Compute the mixed pooling output of the jth neuron in the ith channel in the zth layer

[0035] Among them, σ() is the activation function, p is the width of the pooling kernel, and x z(i,t) is the input of the tth neuron in the ith channel of the zth layer;

[0036] Calculate the output of the fully connected layer z in, is the weight of the line from the ith neuron in the z-1th layer to the jth neuron in the zth layer; is the output value of the i-th neuron in the z-1 layer; is the bias term of the jth neuron in the z-1 layer; is the output value of the i-th neuron in the z-2-th layer.

[0037] Finally, the classification result is calculated, and the final category of the sample is the position of the maximum value among all values.

[0038] 3.3 Input data passes through BiLSTM twice, and Dropout is used to prevent overfitting.

[0039] 3.4 Concatenate the output data of 1DCNN and BiLSTM, and then input the concatenated data into the fully connected layer. The number of neurons in the fully connected layer is set according to the number of categories n.

[0040] 3.5 Based on the number of categories n, determine whether to use the Sigmoid loss function for binary classification or the Softmax loss function for multi-classification.

[0041] 3.6 Output a number representing the category.

[0042] 3.7 Provide classified explanations based on the output results.

[0043] After adopting the present invention, compared with the prior art, the present invention produces obvious beneficial effects:

[0044] (1) Capture data features in depth. The key information loss problem in the convolution calculation of the existing 1DCNN is perfectly solved, and the prediction accuracy of the network model is improved.

[0045] (2) Introduce a more effective information transfer mechanism to better capture complex nonlinear traffic patterns. Aiming at the shortcomings of the existing 1DCNN in nonlinear traffic models, it is proposed to introduce cross-layer information transfer using residual connections to improve the network's expressive power.

[0046] Introducing cross-layer information transmission through residual connections helps the network better capture complex nonlinear traffic patterns and improves the network's expressiveness. DETAILED DESCRIPTION

[0047] The present invention constructs a database abnormal traffic detection system based on the improved 1DCNN algorithm. Figure 1 As shown. The functions of the database abnormal traffic detection system during database use are as follows:

[0048] (1) Data collection: Select the NSL-KDD Dataset and UNSW-NB15 Dataset datasets, which provide raw data (pcap files) about network traffic. These datasets contain normal traffic and different types of abnormal traffic for training and testing models. For example, select a TCP flow record in the NSL-KDD dataset:

[0049] Source IP address: 192.168.1.1

[0050] Destination IP address: 10.0.0.1

[0051] Source port: 12345

[0052] Destination port: 80

[0053] Protocol type: TCP

[0054] Packet size: 1500 bytes

[0055] ...(Other features)

[0056] (2) Data transmission and processing: Utilize distributed processing and asynchronous processing mechanisms to improve the system's processing speed and throughput by distributing data to multiple nodes and processing it asynchronously. Use lightweight data formats, such as Apache Avro or Protocol Buffers, to reduce data transmission overhead. Parse the pcap file and extract relevant features. This may include source IP address, destination IP address, port information, etc. At the same time, the labels need to be processed, marking normal traffic as 0 and abnormal traffic as 1. Preprocess the above TCP traffic record as follows:

[0057] Feature extraction: [192.168.1.1,10.0.0.1,12345,80,TCP,1500,...]

[0058] Label mark: label:0 (normal traffic)

[0059] Data format conversion: [0.2, 0.8, 0.5, 0.3, 0.0, 0.6, ...]

[0060] (3) Detection function: Based on the improved 1DCNN algorithm, the detection function in nonlinear traffic mode, as well as the detection of single traffic sessions and the alarm of malicious events are realized. The model is used to detect each pre-processed data, determine whether it is a malicious event, and provide an alarm and the corresponding probability. Finally, the malicious attack event is persisted.

[0061] Specifically, for the detection algorithm, after determining the number of classifications n, the traffic detection algorithm uses two parts of input data: the first part extracts the spatial characteristics of the traffic through the improved one-dimensional convolutional neural network (1DCNN) model, and the second part extracts the temporal characteristics of the traffic through the bidirectional long short-term memory network (Bidirectional Long Short-Term Memory, BiLSTM) model. The outputs of these two parts of the model are spliced ​​to form a comprehensive feature representation. Finally, through a fully connected layer and the classifier output, the result of classifying the traffic data is obtained. This design can effectively capture spatial and temporal characteristics and improve the accuracy of traffic anomaly detection. The algorithm flow is as follows Figure 2 As shown, the specific algorithm steps are as follows:

[0062] 3.1 Define the number of input categories n.

[0063] 3.2 1DCNN part: Perform one-dimensional convolution operation on the input data. Pool the convolved data. Complete sample classification through Softmax function, and repeat this operation twice.

[0064] This paper proposes a new convolution calculation method, which introduces cross-layer information transmission by residual connection, which helps the network to better capture complex nonlinear traffic patterns and improve the network's expression ability. The specific calculation formula is:

[0065]

[0066] Among them, y z represents the output of the zth layer, ReLU represents the activation function, c z-1 represents the cth channel of the z-1 layer, Attention is the attention mechanism, is the input vector of the z-th layer and the output vector of the i-th channel of the z-1-th layer. Represents the weight matrix of the convolution kernel parameters, where z represents the z-th layer, i represents the i-th channel in the z-th layer, and c represents the c-th channel in the z-1-th layer. represents the bias term b of the i-th channel of the z-th layer, is the output vector of the i-th channel of the z-2-th layer.

[0067] When facing different input feature distributions, single average or maximum pooling may not be able to fully and effectively capture key information. The present invention proposes adaptive hybrid pooling, which makes the model more adaptable to the diversified distribution of input features through dynamic adjustment. When the mean of the local area is large, the method is more inclined to average pooling; when the mean is small, it is more inclined to maximum pooling. This balance helps to retain useful feature information and suppress noise. The hybrid pooling formula is calculated as follows:

[0068]

[0069] Among them, y z(i,j) represents the mixed pooling output of the jth neuron in the ith channel in the zth layer, σ() is the activation function, p represents the width of the pooling kernel, and x z(i,t) Represents the input of the t-th neuron in the ith channel of the z-th layer.

[0070] After multiple rounds of convolution and pooling, the fully connected layer enhances the ability to extract features from the data. The fully connected layer also uses residual connections and stronger nonlinear representation capabilities, which helps improve the performance of the network in complex tasks, improves the performance of the one-dimensional convolutional neural network in database abnormal traffic detection, and enhances the model's expression and generalization capabilities. The formula for the fully connected layer is as follows:

[0071]

[0072] Among them, y z represents the output of the fully connected layer z, n represents the number of neurons in the previous layer of the fully connected layer, i.e., the z-1 layer. represents the weight of the connection from the ith neuron in the z-1th layer to the jth neuron in the zth layer, represents the output value x of the i-th neuron in the z-1 layer, and represents the bias term of the jth neuron in layer z-1, is the output value x of the i-th neuron in the z-2-th layer.

[0073] Finally, the result output needs to use the Softmax function to output the sample classification results. where z i represents the input of the i-th neuron in the output layer, K represents the total number of all categories, and p(x i ) is the probability of the corresponding sample classification on the classifier, and the final category of the sample is the position of the maximum value among all values.

[0074] 3.3 BiLSTM part: Input data passes through a bidirectional long short-term memory network (BiLSTM) twice. In each BiLSTM part, Dropout is used to prevent overfitting.

[0075] 3.4 Concatenate the output data of 1DCNN and BiLSTM. Input the concatenated data into the fully connected layer. The number of neurons in the fully connected layer is set according to the number of categories n. For binary classification, it is set to 1, and for multi-classification, it is set to n.

[0076] 3.5 Based on the number of categories n, determine whether to use the Sigmoid loss function for binary classification or the Softmax loss function for multi-classification.

[0077] 3.6 Output a number representing the category. For binary classification, the output is 0 or 1; for multi-classification, the output is an integer in [0, n-1].

[0078] 3.7 Based on the output results, classify and interpret. In binary classification, 0 represents normal traffic and 1 represents abnormal traffic. In multi-classification, the specific classification is determined based on the mapped labels.

[0079] Assume that the input size of the model is a tensor of n×28×28, where n is the batch size. The convolution layer uses a 3×3 convolution kernel. The output of the convolution part and the output of the BiLSTM part are both 128 in dimension. The outputs of the two are combined to form a tensor of size n×256. This combined tensor passes through a fully connected layer, and the output dimension is 128. Finally, according to the difference between binary classification and multi-classification, Sigmoid and Softmax loss functions are selected to complete the classification explanation. The above TCP traffic record is classified and the classification result label=0 is obtained.

[0080] (4) Visualization function: intuitively display the detection data, including statistics and visualization of traffic data and malicious events, event query and statistical visualization of offline traffic data detection results, query and visualization of past malicious attack events, generate custom reports, and provide customized services to users.

[0081] (5) System management: Use existing identity authentication and permission management mechanisms to ensure system security, including user registration and login, user management, system parameter settings, and other configurations, and provide an easy-to-use user interface.

[0082] The present invention also provides a device, which includes: a data acquisition device, a processor and a memory; the data acquisition device is used to collect data; the memory is used to store one or more program instructions; the processor is used to execute one or more program instructions to execute any of the above methods.

[0083] The present invention further provides a computer-readable storage medium, wherein the computer storage medium contains one or more program instructions, and the one or more program instructions are used to execute any of the above methods.

[0084] The units, devices or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described separately by functions divided into various modules. Of course, when implementing this application, the functions of each module can be implemented in the same or more software and / or hardware, or the modules that implement the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0085] Those skilled in the art also know that, in addition to implementing the controller in a purely computer-readable program code, the controller can be made to implement the same function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered as a hardware component, and the devices for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules for implementing the method and structures within the hardware component.

[0086] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0087] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0088] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0089] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A database abnormal traffic detection method based on an improved 1DCNN network, characterized by: The specific steps of this method are: (1) Network traffic data collection; (2) Transmit and process network traffic data; (3) Adopt the improved 1DCNN algorithm to perform detection under nonlinear traffic mode, and perform single traffic session detection and malicious event alarm; (4) Visual display of test results; (5) System management.

2. According to claim 1, a method for detecting abnormal database traffic based on an improved 1DCNN network is characterized in that: In step (3), the specific algorithm steps are as follows: 3.1 Define the number of input categories n; 3.2 Calculate the output of layer z Among them, ReLU is the activation function; c z-1 is the cth channel of the z-1th layer; Attention is the attention mechanism; is the input vector of the z-th layer, that is, the output vector of the i-th channel of the z-1-th layer; is the weight matrix of the convolution kernel parameters, where z represents the zth layer, i is the i-th channel of the zth layer, and c is the channel in the z-1th layer; is the bias term of the i-th channel of the z-th layer; is the output vector of the i-th channel of the z-2-th layer; Compute the mixed pooling output of the jth neuron in the ith channel in the zth layer Among them, σ() is the activation function, p is the width of the pooling kernel, and x z(i,t) is the input of the tth neuron in the ith channel of the zth layer; Calculate the output of the fully connected layer z in, is the weight of the line from the ith neuron in the z-1th layer to the jth neuron in the zth layer; is the output value of the i-th neuron in the z-1 layer; is the bias term of the jth neuron in the z-1 layer; is the output value of the i-th neuron in the z-2th layer; Finally, the classification result is calculated, and the final category of the sample is the position of the maximum value among all values; 3.3 Input data, passed through BiLSTM twice, and Dropout was used to prevent overfitting; 3.4 Concatenate the output data of 1DCNN and BiLSTM, and then input the concatenated data into the fully connected layer. The number of neurons in the fully connected layer is set according to the number of categories n. 3.5 According to the number of categories n, determine whether to use the Sigmoid loss function for binary classification or the Softmax loss function for multi-classification; 3.6 Output the number representing the category; 3.7 Provide classified explanations based on the output results.

3. A database abnormal traffic detection system based on an improved 1DCNN network, the system comprising: (1) Network traffic data collection module; (2) Network traffic data transmission and processing module; (3) Abnormal traffic detection module; (4) Visual display module of test results; (5) System management module; The feature is that: module (3) uses an improved 1DCNN algorithm to perform detection under nonlinear traffic mode, and performs single traffic session detection and malicious event alarm.

4. According to claim 3, a database abnormal traffic detection system based on an improved 1DCNN network is characterized in that: In module (3), the improved 1DCNN algorithm steps are as follows: 3.1 Define the number of input categories n; 3.2 Calculate the output of layer z Among them, ReLU is the activation function; c z-1 is the cth channel of the z-1th layer; Attention is the attention mechanism; is the input vector of the z-th layer, that is, the output vector of the i-th channel of the z-1-th layer; is the weight matrix of the convolution kernel parameters, where z represents the zth layer, i is the i-th channel of the zth layer, and c is the channel in the z-1th layer; is the bias term of the i-th channel of the z-th layer; is the output vector of the i-th channel of the z-2-th layer; Compute the mixed pooling output of the jth neuron in the ith channel in the zth layer Among them, σ() is the activation function, p is the width of the pooling kernel, and x z(i,t) is the input of the tth neuron in the ith channel of the zth layer; Calculate the output of the fully connected layer z in, is the weight of the line from the ith neuron in the z-1th layer to the jth neuron in the zth layer; is the output value of the i-th neuron in the z-1 layer; is the bias term of the jth neuron in the z-1 layer; is the output value of the i-th neuron in the z-2th layer; Finally, the classification result is calculated, and the final category of the sample is the position of the maximum value among all values; 3.3 Input data, passed through BiLSTM twice, and Dropout was used to prevent overfitting; 3.4 Concatenate the output data of 1DCNN and BiLSTM, and then input the concatenated data into the fully connected layer. The number of neurons in the fully connected layer is set according to the number of categories n. 3.5 According to the number of categories n, determine whether to use the Sigmoid loss function for binary classification or the Softmax loss function for multi-classification; 3.6 Output the number representing the category; 3.7 Provide classified explanations based on the output results.

Citation Information

Patent Citations

  • CNN-GRU hierarchical neural network-based network intrusion detection method

    CN112839034A

  • Network traffic intrusion detection method and device based on deep learning

    CN113660196A

  • Deep space-time hybrid cloud data center network flow real-time detection method

    CN115348074A