CN111310801B - Mixed dimension flow classification method and system based on convolutional neural network - Google Patents
Mixed dimension flow classification method and system based on convolutional neural network Download PDFInfo
- Publication number
- CN111310801B CN111310801B CN202010064381.5A CN202010064381A CN111310801B CN 111310801 B CN111310801 B CN 111310801B CN 202010064381 A CN202010064381 A CN 202010064381A CN 111310801 B CN111310801 B CN 111310801B
- Authority
- CN
- China
- Prior art keywords
- model
- flow
- module
- data
- training
- 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
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 54
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000012549 training Methods 0.000 claims description 60
- 238000011176 pooling Methods 0.000 claims description 50
- 238000012360 testing method Methods 0.000 claims description 29
- 239000011159 matrix material Substances 0.000 claims description 23
- 230000003044 adaptive effect Effects 0.000 claims description 22
- 238000007781 pre-processing Methods 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 12
- 238000013145 classification model Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 10
- 238000006243 chemical reaction Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000007405 data analysis Methods 0.000 claims description 5
- 238000004140 cleaning Methods 0.000 claims description 3
- 230000000295 complement effect Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 5
- 210000002569 neuron Anatomy 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- VYZAMTAEIAYCRO-UHFFFAOYSA-N Chromium Chemical compound [Cr] VYZAMTAEIAYCRO-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000009394 selective breeding Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
- H04L47/2441—Traffic characterised by specific attributes, e.g. priority or QoS relying on flow classification, e.g. using integrated services [IntServ]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a mixed dimension flow classification method and a system based on a convolutional neural network, which relate to the field of computer flow classification.
Description
Technical Field
The invention relates to the field of computer flow classification, in particular to a mixed dimension flow classification method and a system based on a convolutional neural network.
Background
The network traffic classification refers to classifying network communication bidirectional TCP/UDP flows based on TCP/IP protocol according to the application type of the network. The method accurately identifies the network flow information and performs fine management, and is an important basis for realizing effective flow monitoring, network optimization, situation analysis and safety detection. The flow classification is generally divided into a data acquisition stage, a data preprocessing stage, a feature selection stage and a flow classification stage.
The existing feature space construction method adopts artificial selection of proper features for flow data. However, the current network traffic has a wide and complex variety of structures, and the method not only puts higher requirements on the feature selection algorithm, but also increases additional manual calculation cost. The existing flow classification method based on the convolutional neural network can implicitly perform network learning and extract features from training data, avoids the trouble of selecting artificial features, well solves the difference of selecting the features by different classification algorithms, but the flow classification method can only process flow data with fixed length, but the actually collected network flow data is not of fixed length, usually adopts methods such as cutting or compression to process the original data, but the method leads to loss and deformation of gray picture information correspondingly converted to a certain extent, and limits flow classification precision. The existing methods have certain defects, so that the invention of a mixed dimension flow classification method based on a convolutional neural network is necessary to accurately, conveniently and simply classify network flow.
Disclosure of Invention
In order to realize accurate, convenient and simple classification of network traffic, the invention provides a mixed dimension traffic classification method and a system based on a convolutional neural network, and the specific technical scheme is as follows:
a mixed dimension flow classification method based on a convolutional neural network comprises the following steps:
step one: collecting flow data of a computer through a network protocol data analysis tool, marking a corresponding application type flow label on each flow data, and writing each flow data into a txt file;
step two: reading the txt file generated in the first step through a computer programming language, reading the actual length of each piece of flow data, then opening and squaring the actual length of the flow data, adding 1 after omitting decimal places from the square opening result, and converting the decimal places into a dimension value of a two-dimensional matrix as one-dimensional flow data, wherein the complement value of a blank bit is 0, so as to obtain the two-dimensional flow matrix;
step three: preprocessing the two-dimensional flow matrix data obtained by the conversion in the step two through a normalization algorithm to finish gray level image mapping, and dividing the gray level image into a test data set and a training data set according to the proportion of 3:7;
step four: replacing a general pooling layer connected with the full-connection layer in the convolutional neural network model-Lenet-5 model with a self-Adaptive maximum pooling layer, and outputting after replacement to obtain an Adaptive-CNN model;
step five: the output dimension of the Adaptive maximum pooling layer obtained through the replacement in the fourth step is fixed, and the dimension size of the convolution kernel of the Adaptive maximum pooling layer in the Adaptive-CNN model and the moving step length of the convolution kernel are adjusted according to different dimension sizes of input flow data;
step six: reading gray level pictures in the training data set in the third step, importing the gray level pictures into the adjusted Adaptive-CNN model in the fifth step for training, and outputting a trained flow classification model;
step seven: and D, reading the gray level pictures in the test data set in the third step, importing the flow classification model trained in the sixth step, and outputting the classification precision of the test data set.
Further, the network protocol data analysis tool in the first step adopts free network traffic real-time monitoring software Microsoft Network Monitor issued by microsoft, and the Microsoft Network Monitor is installed on a computer.
Further, the Adaptive-CNN model output in the fourth step is to replace a general pooling layer connected with a full-connection layer in a convolutional neural network model-Lenet-5 model with an Adaptive maximum pooling layer, and adjust the size of a convolution kernel and the moving step length of the convolution kernel in the Adaptive maximum pooling layer by fixing the output data dimension of the Adaptive maximum pooling layer so as to realize classification of flow data with different dimensions.
The invention also comprises a mixed dimension flow classification system based on the convolutional neural network, which comprises a data acquisition module, a data preprocessing module, a model construction module, a model training module, a model optimization module and a flow classification module;
the input end of the data acquisition module is connected with a computer, the output end of the data acquisition module is connected with the data preprocessing module, the model building module is connected with the model training module, the model training module is connected with the model optimizing module, and the model optimizing module is connected with the flow classifying module;
the data acquisition module is used for acquiring one-dimensional network flow data of the computer, cleaning the acquired one-dimensional network flow data of the computer and marking corresponding application labels;
the data preprocessing module is used for converting one-dimensional network flow data into a two-dimensional gray level picture, and obtaining a test data set and a training data set after conversion, wherein the test data set is imported into the flow classification module, and the training data set is imported into the model training module;
the model construction module is used for selecting a convolutional neural network model and importing the selected convolutional neural network model into the model training module;
the model training module is used for training the training data set through a convolutional neural network model, and outputting a training model to the model optimizing module after training;
the model optimization module is used for optimizing and adjusting a convolutional neural network model output by the model training module, and adjusting the output dimension, the learning rate and the normalization parameters of the Adaptive pooling layer of the Adaptive-CNN model according to the classification precision of the training data set;
the flow classification module is used for testing the data of the test data set in the trained flow classification model imported by the test data set and outputting the classification precision of the test data set.
The beneficial effects of the invention are as follows:
the method aims at the problem that the length of actually acquired flow data is not fixed, and the convolutional neural network-Lenet-5 model is partially modified, so that the information loss and deformation caused by manually selecting the cutting direction or the compression range are reduced, the expression level of the characteristics is increased, the influence of human factors is reduced, and the accuracy of flow classification is improved.
Drawings
Fig. 1 is a schematic flow chart of a mixed dimension flow classification method based on a convolutional neural network according to an embodiment of the present invention;
fig. 2 is a schematic diagram of network traffic data acquisition according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of an adaptive max-pooling layer according to an embodiment of the present invention.
Fig. 4 is a schematic structural diagram of a mixed dimension flow classification system based on a convolutional neural network according to an embodiment of the present invention.
Detailed Description
For a better understanding of the present invention, reference is made to the following description of the invention, taken in conjunction with the accompanying drawings and specific examples:
as shown in fig. 1 and 2, a method for classifying mixed dimension flow based on convolutional neural network includes the following steps:
step one: the method comprises the steps of collecting flow data of 8 desktops connected to a campus network by using a free network flow real-time network protocol data analysis tool Microsoft Network Monitor issued by Microsoft, marking corresponding application type flow labels on each flow data, and writing each flow data into txt files.
The traffic data of 8 desktops accessing the campus network comprises traffic data of 11 common application software in the campus network traffic, and the 11 common application software comprises webpage classes (Chrome browser, fireFox browser, IE browser and 360 browser), communication classes (QQ), download classes (Thunder and BaiduNetDisk) and video classes (IQIYI, mgtv, YOUKU and QQLive) for quantitative data acquisition.
Taking grabbing application software QQ as an example, the specific operation is as follows:
Conversation.ProcessName＝＝"QQ.exe"and
Ipv4.TotalLength>＝900 and
Ipv4.TotalLength<1444
the first row represents the acquired application type, and the second two rows are obtained by analyzing the random grabbing flow data, and the data size of most flow data is found to be evenly distributed between 900-1444 bytes, so that the range of acquired data is set to be more than 900 bytes and less than 1444 bytes.
Microsoft Network Monitor and Microsoft Network Monitor are installed on the 8 desktops connected to the campus network, so that various network interface devices on the 8 desktops connected to the campus network are displayed in the same picture, real-time inflow or outflow flow of each interface is displayed, flow data of inflow or outflow is updated once every second Microsoft Network Monitor, and a user can master the most real-time network flow information at any time.
Step two: reading the txt file generated in the first step through a computer programming language, reading the actual length of each piece of flow data, then opening and squaring the actual length of the flow data, adding 1 after omitting decimal places from the square opening result, and converting the decimal places into a dimension value of a two-dimensional matrix as one-dimensional flow data, wherein the complement value of a blank bit is 0, so as to obtain the two-dimensional flow matrix;
if the square root value is not an integer, adding 1 after rounding the square root, and converting the square root value into a dimension value of a two-dimensional matrix as one-dimensional flow data;
suppose A 11 For one piece of traffic data, it can be expressed as:
wherein a is 11 ～a nn For 16 bytes, converted into matrix M in picture format 11 The method comprises the following steps:
wherein b 11 ＝a 11 ×16+a 12 And so on;
matrix M after conversion 11 Carrying out normalization processing to finish conversion from the two-dimensional matrix to the gray level picture;
said normalization process being M 11 ＝M 11 /255。
Step three: preprocessing the two-dimensional matrix data obtained by the conversion in the step two through a normalization algorithm to finish gray level picture mapping, and taking the gray level picture as 3:7 is divided into a test data set and a training data set, and is specifically as follows:
the gray picture data set is assumed to be represented as a two-dimensional matrix T, wherein A-K respectively represent application types of different acquired network flow data, 1-i are flow data which respectively correspond to the same application type in the acquired flow data and have different dimension values of the two-dimensional matrix, n is the total number of the flow data with the dimension of i of the two-dimensional matrix in the application types A-K, and the value of the total number is not unique.
The concrete steps are as follows:
then A 11 The method comprises the following steps:
wherein a is 11 ～a nn For 16 bytes, converting it into picture format, i.e
Wherein b 11 ＝a 11 ×16+a 12 And so on.
The matrix can be expressed as:
wherein AA is 1i Representing a traffic dataset of application type a and dimension i.
Each row is normalized as follows to obtain N 1 ：
N 1 ＝[AA 11 ,AA 12 ,AA 13 …AA 1i ]/255
The quantized matrix can be expressed as:
T'＝[N 1 ,N 2 ,N 3 …N k ]。
step four: and replacing a general pooling layer connected with the full-connection layer in the convolutional neural network model-Lenet-5 model with a self-Adaptive maximum pooling layer, outputting the self-Adaptive maximum pooling layer after replacement to obtain an Adaptive-CNN model, wherein the output Adaptive-CNN model is obtained by replacing the general pooling layer connected with the full-connection layer in the convolutional neural network model-Lenet-5 model with the self-Adaptive maximum pooling layer, and adjusting the size of a convolution kernel and the moving step length of the convolution kernel in the self-Adaptive maximum pooling layer by fixing the dimension of output data of the self-Adaptive maximum pooling layer.
As shown in fig. 3, the pooling layer is divided into four types, respectively:
type one: general pooling, i.e. maximum pooling (MaxPooling) and average pooling (AveragePooling). The pooling layer in the Lenet-5 model employs a generic pooling layer.
Type two: global pooling, global max pooling (globalpaxpoling) and global average pooling (globalaneragepooling).
Type three: adaptive pooling, i.e. adaptive maximum pooling (adaptive maxpooling) and adaptive average pooling (adaptive average pooling).
Type four: pyramid pooling, consisting of two or more generic pooling layers, can be divided into pyramid maximum pooling and pyramid average pooling.
The four types of pooling layers are in the form of:
wherein,an input value representing a jth neuron of a layer l, β representing a network multiplicative parameter of the neuron, b representing a bias of the neuron, down (·) representing a subsampling function that is a weighted sum of each n×n-sized region in the input image, so that the size of the output image becomes +_of the input image size>f represents an activation function, such as Sigmoid function, tanh function, reLU function, etc,/->Representing the connection weights of the j-th neuron of the layer 1.
Step five: the output dimension of the Adaptive maximum pooling layer obtained through the replacement in the fourth step is fixed, and the dimension size of the convolution kernel of the Adaptive maximum pooling layer in the Adaptive-CNN model and the moving step length of the convolution kernel are adjusted according to different dimension sizes of input flow data, wherein the steps are as follows:
the self-adaptive pooling layer is used for fixing the output dimension of the pooling layer, automatically adjusting the size of the convolution kernel and the moving step length of the convolution kernel, and specifically comprises the following steps:
stride＝floor(input_size/output_size)
kernel_size＝input_size-(output_size-1)×stride
where stride is the moving step of the convolution kernel, input_size is the input data dimension of the adaptive pooling layer, output_size is the artificially set output data dimension, kernel_size is the convolution kernel size, and floor is the rounding down.
Step six: and D, reading gray level pictures in the training data set in the third step, importing the gray level pictures into the adjusted Adaptive-CNN model in the fifth step for training, adjusting the learning rate and normalization of the Adaptive-CNN model according to the training result, namely the accuracy of classification per rotation, taking a model with more stable classification effect and better classification accuracy as a flow classification model which is finally trained by the module, and outputting the trained flow classification model.
Step seven: and D, reading the gray level pictures in the test data set in the third step, importing the flow classification model trained in the sixth step, outputting the classification precision of the test data set, optimizing the test data set, adjusting the learning rate or normalization parameters of the training model according to the output classification precision, obtaining an optimal classification model, and outputting the optimal classification precision.
The model optimization module adjusts the learning rate of the training model or the basis of the normalization parameters according to the output classification precision as follows:
if the classification accuracy falls into local optimum, namely the optimal value of the classification accuracy of the training model maintains the classification accuracy of a certain round for a long time, the model optimization module adjusts the normalization parameters of the training model.
If the classification accuracy changes slowly, namely the classification accuracy of the training model always presents a small-amplitude rising and falling trend or a small-amplitude fluctuation state, the model optimization module adjusts the learning rate of the training model.
As shown in FIG. 4, the mixed dimension flow classification system based on the convolutional neural network comprises a data acquisition module, a data preprocessing module, a model construction module, a model training module, a model optimization module and a flow classification module.
The input end of the data acquisition module is connected with a computer, the output end of the data acquisition module is connected with the data preprocessing module, the model building module is connected with the model training module, the model training module is connected with the model optimizing module, and the model optimizing module is connected with the flow classifying module.
The data acquisition module is used for acquiring one-dimensional network flow data of the computer, cleaning the acquired one-dimensional network flow data of the computer and marking corresponding application labels.
The data preprocessing module is used for converting one-dimensional network flow data into two-dimensional gray level pictures, obtaining a test data set and a training data set after conversion, leading the test data set into the flow classification module, and leading the training data set into the model training module.
The model construction module is used for selecting the convolutional neural network model and importing the selected convolutional neural network model into the model training module.
The model training module is used for training the training data set through the convolutional neural network model, and outputting the training model to the model optimizing module after training.
The model optimization module is used for optimizing and adjusting the convolutional neural network model output by the model training module, and adjusting the output dimension, the learning rate and the normalization parameters of the Adaptive pooling layer of the Adaptive-CNN model according to the classification accuracy of the training data set.
The flow classification module is used for testing the data of the test data set in the trained flow classification model imported by the test data set and outputting the classification precision of the test data set.
Based on the given model, the public traffic dataset Moore dataset shown in table 1 and the dataset constructed using step one of the systems given herein shown in table 2 were tested, respectively.
Table 1 Moore flow data statistics table
TABLE 2 actual network flow data distribution Table (Unit/strip)
TABLE 3 Overall accuracy and test timetable with different algorithms
As shown in table 3, for Moore dataset, the Adaptive-CNN model not only reduces workload, but also improves classification accuracy of traffic, compared with the traffic classification performed after feature selection of traffic data through PCA (Principal Components Analysis) and SRP (Sparse Random Projection), respectively.
Aiming at the data set constructed by the invention, the Adaptive-CNN model can directly process the flow data without manually extracting flow characteristics, can realize the classification of mixed flow data, can classify the flow data with different dimensions by adopting the same training model, and reduces the influence of human factors on classification results. The classification accuracy of the Adaptive-CNN model is up to 87.25%.
The present invention is not limited to the above embodiments, but is to be accorded the widest scope consistent with the principles and other features of the present invention.
Claims (4)
1. The mixed dimension flow classification method based on the convolutional neural network is characterized by comprising the following steps of:
step one: collecting flow data of a computer through a network protocol data analysis tool, marking a corresponding application type flow label on each flow data, writing each flow data into a txt file, and updating the flow data flowing in or out once every second;
step two: reading the txt file generated in the first step through a computer programming language, reading the actual length of each piece of flow data, then opening and squaring the actual length of the flow data, adding 1 after omitting decimal places from the square opening result, and converting the decimal places into a dimension value of a two-dimensional matrix as one-dimensional flow data, wherein the complement value of a blank bit is 0, so as to obtain the two-dimensional flow matrix;
step three: preprocessing the two-dimensional flow matrix data obtained by the second conversion step through a normalization algorithm to finish gray level image mapping, dividing gray level images into a test data set and a training data set according to a certain proportion, and specifically, expressing the gray level image data set as a two-dimensional matrix T, wherein A-K respectively represent application types of different acquired network flow data, 1-i respectively correspond to the flow data with the same application type in the acquired flow data, the two-dimensional matrix has different dimension values, and n is the total number of the flow data with the dimension of i in the two-dimensional matrix in the application types A-K, and the values are not unique;
the concrete steps are as follows:
then A 11 The method comprises the following steps:
wherein a is 11 ～a nn For 16 bytes, converting it into picture format, i.e
Wherein b 11 ＝a 11 ×16+a 12 And so on;
the matrix can be expressed as:
wherein AA is 1i A flow data set with a dimension i and an application type A is represented;
each row is normalized as follows to obtain N 1 ：
N 1 ＝[AA 11 ,AA 12 ,AA 13 …AA 1i ]255
The quantized matrix is expressed as:
T'＝[N 1 ,N 2 ,N 3 …N k ]；
step four: replacing a general pooling layer connected with the full-connection layer in the convolutional neural network model-Lenet-5 model with a self-Adaptive maximum pooling layer, and outputting after replacement to obtain an Adaptive-CNN model;
step five: the output dimension of the Adaptive maximum pooling layer obtained through the replacement in the fourth step is fixed, and the dimension size of the convolution kernel of the Adaptive maximum pooling layer in the Adaptive-CNN model and the moving step length of the convolution kernel are adjusted according to different dimension sizes of input flow data;
step six: reading gray level pictures in the training data set in the third step, importing the gray level pictures into the adjusted Adaptive-CNN model in the fifth step for training, and outputting a trained flow classification model;
step seven: and D, reading the gray level pictures in the test data set in the third step, importing the flow classification model trained in the sixth step, and outputting the classification precision of the test data set.
2. The method for classifying mixed dimension traffic based on convolutional neural network according to claim 1, wherein the network protocol data analysis tool in the first step adopts a free network traffic real-time monitoring software Microsoft Network Monitor issued by microsoft, and the Microsoft Network Monitor is installed on a computer.
3. The method for classifying the mixed dimension flow based on the convolutional neural network according to claim 1, wherein the Adaptive-CNN model output in the fourth step replaces a general pooling layer connected with a full-connection layer in a convolutional neural network model-Lenet-5 model with an Adaptive maximum pooling layer, and the size of a convolution kernel and the moving step length of the convolution kernel in the Adaptive maximum pooling layer are adjusted by fixing the output data dimension of the Adaptive maximum pooling layer so as to classify the flow data with different dimensions.
4. The mixed dimension flow classification system based on the convolutional neural network is characterized by comprising a data acquisition module, a data preprocessing module, a model construction module, a model training module, a model optimization module and a flow classification module;
the input end of the data acquisition module is connected with a computer, the output end of the data acquisition module is connected with the data preprocessing module, the model building module is connected with the model training module, the model training module is connected with the model optimizing module, and the model optimizing module is connected with the flow classifying module;
the data acquisition module is used for acquiring one-dimensional network flow data of the computer, cleaning the acquired one-dimensional network flow data of the computer and marking corresponding application labels;
the data preprocessing module is used for converting one-dimensional network flow data into a two-dimensional gray level picture, and obtaining a test data set and a training data set after conversion, wherein the test data set is imported into the flow classification module, and the training data set is imported into the model training module;
the model construction module is used for selecting a convolutional neural network model and importing the selected convolutional neural network model into the model training module;
the model training module is used for training the training data set through a convolutional neural network model, and outputting a training model to the model optimizing module after training;
the model optimization module is used for optimizing and adjusting a convolutional neural network model output by the model training module, and adjusting the output dimension, the learning rate and the normalization parameters of the Adaptive pooling layer of the Adaptive-CNN model according to the classification precision of the training data set;
the flow classification module is used for testing the data of the test data set in the trained flow classification model imported by the test data set and outputting the classification precision of the test data set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010064381.5A CN111310801B (en) | 2020-01-20 | 2020-01-20 | Mixed dimension flow classification method and system based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010064381.5A CN111310801B (en) | 2020-01-20 | 2020-01-20 | Mixed dimension flow classification method and system based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111310801A CN111310801A (en) | 2020-06-19 |
| CN111310801B true CN111310801B (en) | 2024-02-02 |
Family
ID=71148969
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010064381.5A Active CN111310801B (en) | 2020-01-20 | 2020-01-20 | Mixed dimension flow classification method and system based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111310801B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111817981A (en) * | 2020-07-01 | 2020-10-23 | 黄东 | Network traffic classification method based on deep learning |
| CN114726741A (en) * | 2020-12-18 | 2022-07-08 | 中国科学院沈阳自动化研究所 | Edge network state perception modeling method based on characterization learning |
| CN112906019B (en) * | 2021-02-04 | 2022-11-11 | 南京邮电大学 | Flow data generation method, device and system based on improved DCGAN model |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107194559B (en) * | 2017-05-12 | 2020-06-05 | 杭州电子科技大学 | Workflow identification method based on three-dimensional convolutional neural network |
| CN108932480B (en) * | 2018-06-08 | 2022-03-15 | 电子科技大学 | Distributed optical fiber sensing signal feature learning and classifying method based on 1D-CNN |
| CN109033169B (en) * | 2018-06-21 | 2021-08-10 | 东南大学 | Mobile traffic classification method based on multistage weight conversion and convolutional neural network |
| CN109361617B (en) * | 2018-09-26 | 2022-09-27 | 中国科学院计算机网络信息中心 | Convolutional neural network traffic classification method and system based on network packet load |
| CN109766992B (en) * | 2018-12-06 | 2020-12-04 | 北京工业大学 | Industrial control abnormity detection and attack classification method based on deep learning |
| CN109840279A (en) * | 2019-01-10 | 2019-06-04 | 山东亿云信息技术有限公司 | File classification method based on convolution loop neural network |
| CN110009579B (en) * | 2019-03-14 | 2020-11-24 | 桂林航天工业学院 | Image restoration method and system based on brain storm optimization algorithm |
-
2020
- 2020-01-20 CN CN202010064381.5A patent/CN111310801B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111310801A (en) | 2020-06-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111310801B (en) | Mixed dimension flow classification method and system based on convolutional neural network | |
| KR102173610B1 (en) | Vehicle license plate classification method, system, electronic device and media based on deep learning | |
| US20210182611A1 (en) | Training data acquisition method and device, server and storage medium | |
| CN110059694B (en) | Intelligent identification method for character data in complex scene of power industry | |
| CN108563739B (en) | Weather data acquisition method and device, computer device and readable storage medium | |
| CN112232371B (en) | American license plate recognition method based on YOLOv3 and text recognition | |
| WO2022095376A1 (en) | Aspect-based sentiment classification method and apparatus, device, and readable storage medium | |
| CN106971198A (en) | A kind of pneumoconiosis grade decision method and system based on deep learning | |
| CN116760772B (en) | Control system and method for converging flow divider | |
| CN112949704A (en) | Tobacco leaf maturity state identification method and device based on image analysis | |
| CN108121781A (en) | Search method of related feedback images with parameter optimization is chosen based on efficient sample | |
| CN114896067B (en) | Automatic generation method and device of task request information, computer equipment and medium | |
| CN113065609B (en) | Image classification method, device, electronic equipment and readable storage medium | |
| CN114067119A (en) | Training method of panorama segmentation model, panorama segmentation method and device | |
| CN118172935B (en) | Intelligent high-speed management system and method based on digital twinning | |
| CN111159241B (en) | Click conversion estimation method and device | |
| CN114580517A (en) | Method and device for determining image recognition model | |
| CN112085668B (en) | Image tone mapping method based on region self-adaptive self-supervision learning | |
| CN117768338A (en) | Network traffic matrix prediction method based on gradient lifting tree | |
| CN115905983A (en) | Artificial intelligence data classification system | |
| CN114332129B (en) | Model training method and device, image segmentation method and device | |
| CN112860531B (en) | Block chain wide consensus performance evaluation method based on deep heterogeneous graph neural network | |
| WO2023069085A1 (en) | Systems and methods for hand image synthesis | |
| CN113610111B (en) | Fusion method, device, equipment and storage medium of distributed multi-source data | |
| CN117218396B (en) | Automatic visual sample data labeling method and system based on large model |
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 |