CN115659323A - Intrusion detection method based on information entropy theory and convolution neural network - Google Patents
Intrusion detection method based on information entropy theory and convolution neural network Download PDFInfo
- Publication number
- CN115659323A CN115659323A CN202211094816.6A CN202211094816A CN115659323A CN 115659323 A CN115659323 A CN 115659323A CN 202211094816 A CN202211094816 A CN 202211094816A CN 115659323 A CN115659323 A CN 115659323A
- Authority
- CN
- China
- Prior art keywords
- data
- neural network
- convolutional neural
- intrusion detection
- information entropy
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 40
- 238000013528 artificial neural network Methods 0.000 title description 4
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 41
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000007637 random forest analysis Methods 0.000 claims abstract description 9
- 238000011156 evaluation Methods 0.000 claims abstract description 3
- 238000007781 pre-processing Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 2
- 230000006399 behavior Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 abstract description 2
- 238000000605 extraction Methods 0.000 abstract description 2
- 230000003111 delayed effect Effects 0.000 abstract 2
- 230000004913 activation Effects 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 4
- 238000011176 pooling Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000011157 data evaluation Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
An intrusion detection method based on the combination of an information entropy theory and a convolutional neural network comprises the steps of firstly, converting character type data into numerical type data, standardizing the data and normalizing the data; and then, putting the data set into a convolutional neural network for dimensionality reduction and classification, calculating by combining information entropy uncertainty, and performing delayed relearning classification decision on part of data, wherein a random forest method is selected by a delayed decision method. When the intrusion behavior occurs, normal data and attack data can be distinguished by using the trained model. The method utilizes the characteristic of superior performance on the characteristic extraction capability and the classification learning effect of the convolutional neural network and combines the information entropy theory to evaluate the classified data, and the evaluation result is used as a secondary learning classification decision basis.
Description
Technical Field
The invention relates to an intrusion detection method based on the combination of an information entropy theory and a convolutional neural network, and belongs to the technical field of intrusion detection in networks.
Background
In recent years, with the continuous change of network technology and the continuous expansion of network scale, network security events at home and abroad are frequent, and the problem of network security is more emphasized. Research into intrusion detection systems has become an important part of current network security development. The research of the intrusion detection system is to make up the deficiency of attack defense in the traditional firewall and strengthen the monitoring of the network and the system operation condition, and discover various attack attempts, attack behaviors or attack results as much as possible to ensure the confidentiality, integrity and availability of network system resources. Intrusion detection systems have been developed for decades and have achieved some success by now, but there is much room for research on intrusion detection systems.
The conventional intrusion detection system still has some problems, specifically: the attack behavior can not be automatically researched and distinguished, the response is not timely enough when the detection is carried out on a large data scale, and the signature database needs to be updated in real time. The above problems will cause the system to have a low accuracy of detecting the data flow and generate a relatively high false alarm rate during the detection. When the unknown abnormal flow is faced, a better division mode is not provided, and the autonomous learning ability is lacked.
Disclosure of Invention
In order to make up for the defects of the prior art, the invention provides an intrusion detection method based on the combination of an information entropy theory and a convolutional neural network. The accuracy of detection is improved through the excellent performance of feature extraction and classification of the convolutional neural network, a belief theory is introduced to serve as secondary judgment, the occurrence of misjudgment is reduced, and the false alarm rate is reduced, so that the problems in the background art are solved.
In order to achieve the purpose, the invention provides the following technical scheme: an intrusion detection method based on the combination of an information entropy theory and a convolutional neural network comprises the following specific steps:
the first step is as follows: acquiring a training data set and a testing data set from the data set to perform data preprocessing on the data set;
the second step: transmitting the training set after data preprocessing into a network model for training to obtain a trained network model;
the third step: calculating an information entropy value for evaluation through an information entropy theory by obtaining probability distribution and confidence of data through a convolution bible network model;
the fourth step: directly outputting a classification result by the convolutional neural network for the data with the entropy value smaller than the threshold value, and learning and reclassifying the data larger than the threshold value by a random forest model;
the fifth step: and transmitting the test set after data preprocessing into the trained network model to obtain a classification result.
As a preferred scheme, the first step of performing data preprocessing on the intrusion detection data set specifically comprises the following steps:
(1) Because some characteristics in the intrusion detection data set are character type data, the character type data needs to be converted into numerical type data;
(2) In order to reduce the influence of high data dispersity and the numerical value on the model in the characteristics, the converted numerical data needs to be subjected to standardization processing;
(3) And carrying out data normalization on the normalized data in order to reduce the calculated amount of the model, so that the data are mapped into a [0,1] interval.
As a preferred scheme, the second step trains the network model with the processed network data, and the specific process of obtaining the trained network model is as follows:
(1) Calculating the output value of each neuron in a forward direction;
(2) Determining an optimization objective function;
(3) Carrying out forward and backward propagation to update network weight parameters according to gradient guidance of a convolutional neural network loss function;
(4) And repeating the three steps until the network error is less than a given value, and determining an optimal convolutional neural network model.
As a preferred scheme, the specific process of calculating the information entropy value by the information entropy theory to evaluate the probability distribution and the confidence of the data obtained in the third step is as follows:
(1) Carrying out entropy calculation on probability distribution output by the convolutional neural network through an information entropy theory;
(2) And (4) comparing the entropy value with a threshold value to further divide data needing secondary learning classification.
As a preferred scheme, the fourth step directly outputs the classification result from the convolutional neural network for the data with the entropy value smaller than the threshold, and the specific process of learning and reclassifying the data larger than the threshold by the random forest model is as follows;
(1) The convolutional neural network outputs a data category smaller than a threshold value, and the random forest further learns and divides data higher than the threshold value;
(2) Counting the uncertain sample rate in secondary classification;
(3) And combining and outputting the primary classification result and the secondary learning classification result.
As a preferred scheme, the specific process of transmitting the test set after data preprocessing into a trained network model to obtain a classification result is as follows;
(1) Performing parameter optimization on the intrusion detection model by a training set, and obtaining an optimal intrusion detection model when the uncertain sample rate of secondary classification reaches the lowest through threshold adjustment;
(2) And inputting a test set to test the intrusion detection model to obtain a final classification result.
Compared with the prior art, the invention provides an intrusion detection method based on the information entropy theory and the convolution neural network, which has the following beneficial effects:
the invention adopts the convolutional neural network with deep learning effective algorithm to extract large-scale data features and excellent classification capability, and applies the convolutional neural network to intrusion detection to improve the accuracy of the intrusion detection. And further evaluating and dividing the data classes by combining the confidence degrees generated by classifying the convolutional neural network with the information entropy theory. And the data types with large entropy and high uncertainty are relearned, so that the false alarm rate of intrusion detection is reduced. Thereby improving the overall intrusion detection performance.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a diagram of a classification method for a CNN model;
fig. 3 is a diagram of a RF model classification method.
Detailed Description
In order to more clearly understand the technical features, objects, and effects of the present invention, embodiments of the present invention will now be described with reference to the accompanying drawings.
As shown in fig. 1, an intrusion detection method based on the information entropy theory in combination with the convolutional neural network specifically includes the following steps:
the method comprises the following steps: carrying out data preprocessing on the intrusion detection data set;
the data preprocessing is divided into 3 steps:
1) Conversion of character-type data into numerical-type data
Because some features in the intrusion detection dataset are character-type data, and the convolutional neural network can only process numerical data, the character-type data in the dataset is first converted into numerical data.
2) Data standardization
In order to reduce the influence of high data dispersity and numerical value on a convolutional neural network in the features, the data converted into numerical values are standardized, the average value and the average absolute error of each characteristic value are calculated, and the formula is as follows:
wherein x is k Means, S, representing the k-th attribute k Mean absolute error, x, representing the kth feature ik Representing the kth attribute of the ith record. Then, a normalization metric is performed on each data record, and the formula is as follows:
wherein, Z ik K attribute value representing normalized ith data record。
3) Data normalization
And normalizing the normalized data to ensure that the numerical value of the data is in a [0,1] interval so as to reduce the calculated amount of the model, wherein the normalization processing formula is as follows:
step two: as shown in fig. 2, preprocessing the data and then transmitting the preprocessed data into a convolutional neural network model for training to obtain a trained network model;
training and tuning the convolutional neural network comprises the following three stages:
1) And (3) performing convolution on the preprocessed data, obtaining mapping of corresponding positions after each position is passed, and activating a function to form a new characteristic matrix through network parameter setting when a complete characteristic matrix is traversed. When all feature samples complete the feature mapping, the stack combination forms the feature map of the convolution. The convolution calculation formula is as follows:
m is the set of input samples and M is,
is the feature matrix output from the previous layer, "+" is the convolution operation, n is the current layer number,
is a convolution kernel, b n And selecting a Relu function as the activation function for the offset phi. Compared with other activation functions, the unsaturated characteristic of the Relu activation function can effectively avoid the gradient disappearance phenomenon.
2) The features output by the convolutional layer are sampled by the pooling layer, and all features are summarized by using data features of probability statistics. Not only can reduce the dimension, but also can ensureLeaving the most valid information. The method selects a maximum pooling method.
For the pooled feature matrix, the pooling method has the following calculation formula:
3) The convolution neural network with updated network parameters can obtain an original characteristic output value through forward propagation, the network parameters need to be updated through backward propagation, and the loss function of the convolution neural network is as follows:
wherein y is the actual value of the value,
is a predicted value. The parameter values are updated once per training. When the network trains the data sample, the iterative update formula for updating the weight ω and the bias b in the process of reducing the network loss is as follows:
alpha is learning rate
Is the partial derivative of the weight w and the bias b by the loss function.
Step three: the CNN network output layer obtains probability distribution values (P) of two types of data through a softmax activation function 1 ,P 2 ) Probability of passing the resulting data trafficThe distribution is subjected to entropy calculation, and the data is further screened and evaluated through an information entropy theory; .
The data evaluation method is as follows:
by obtaining the probability distribution value (P) of two types of data 1 ,P 2 ) The calculation formula for calculating the information entropy is as follows:
the entropy represented by H is also expressed as the uncertainty of the data, and n =2,P is obtained by classifying the data into two types i Is the probability distribution of two predicted results and can also represent the confidence of data in two categories. The CNN network output layer obtains probability distribution values (P) of two types of data through a softmax activation function 1 ,P 2 ). And (4) rapidly sampling the probability distribution of all samples to calculate the uncertainty, wherein the calculated result reflects the size of the uncertainty of the data. The information entropy calculation formula of n =2 is as follows:
H＝-(P 1 log 2 P 1 +P 2 log 2 P 2 )
when the classified data is (P) 1 ,P 2 ) The more the probability distribution is even, the larger the calculated H is, the higher the uncertainty of the classification sample is, and vice versa (P) 1 ,P 2 ) The larger the difference in probability values, the lower the uncertainty of the classified sample. Through the division of H, the samples can be divided into samples with high uncertainty and samples with low uncertainty, and the uncertainty of the obtained data samples can be used as a judgment basis for delay classification decision.
Step four: directly outputting a classification result by the convolutional neural network for the data with the entropy value smaller than the threshold theta, learning and reclassifying the data larger than the threshold theta by using a random forest model shown in figure 3, and counting the uncertain sample rate during secondary classification;
step five: obtaining an optimal intrusion detection model when the uncertain sample rate of secondary classification reaches the lowest through adjusting the threshold value, and inputting test data in the trained optimal model to obtain a classification result;
the method comprises the steps of firstly, transmitting data into a convolutional neural network for classification through a data preprocessing stage, carrying out information entropy calculation through the confidence coefficient of the data generated by CNN, directly outputting a classification result by the CNN when the entropy value is smaller than a threshold value theta, and further classifying by a random forest model when the entropy value is larger than the threshold value theta. And combining the classification results obtained by the two as common output. The overall performance of the model can be improved, and the risk of model misclassification is reduced.
While the present invention has been described in detail with reference to the embodiments, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.
Claims (6)
1. An intrusion detection method based on the combination of an information entropy theory and a convolutional neural network is characterized in that: the method comprises the following specific steps:
the first step is as follows: acquiring a training data set and a testing data set from the data set to perform data preprocessing on the data set;
the second step is that: transmitting the training set after data preprocessing into a network model for training to obtain a trained network model;
the third step: calculating an information entropy value for evaluation through an information entropy theory by obtaining probability distribution and confidence of data through a convolution bible network model;
the fourth step: directly outputting a classification result by the convolutional neural network for the data with the entropy value smaller than the threshold value, and learning and reclassifying the data larger than the threshold value by a random forest model;
the fifth step: and transmitting the test set after data preprocessing into the trained network model to obtain a classification result.
2. The intrusion detection method based on the information entropy theory combined with the convolutional neural network as claimed in claim 1, wherein: the first step is to perform data preprocessing on the intrusion detection data set, and the specific process is as follows:
(1) Because some characteristics in the intrusion detection data set are character type data, the character type data needs to be converted into numerical type data;
(2) In order to reduce the influence of high data dispersity and the influence of the numerical value on the model in the characteristics, the converted numerical data needs to be subjected to standardization processing;
(3) And normalizing the data after the normalization processing to reduce the calculation amount of the model, so that the data are mapped into a [0,1] interval.
3. The intrusion detection method based on the information entropy theory combined with the convolutional neural network as claimed in claim 1, wherein: in the second step, the network model is trained by the processed network data, and the specific process of obtaining the trained network model is as follows:
(1) Calculating the output value of each neuron in a forward direction;
(2) Determining an optimization objective function;
(3) Carrying out forward and backward propagation to update network weight parameters according to gradient guidance of a convolutional neural network loss function;
(4) And repeating the three steps until the network error is smaller than a given value, and determining the optimal convolutional neural network model.
4. The intrusion detection method based on the information entropy theory combined with the convolutional neural network as claimed in claim 1, wherein: the third step calculates the information entropy value through the information entropy theory to evaluate the probability distribution and the confidence coefficient of the obtained data, and the specific process is as follows:
(1) Carrying out entropy calculation on probability distribution output by the convolutional neural network through an information entropy theory;
(2) And dividing data needing secondary learning classification by comparing the entropy value with a threshold value.
5. The intrusion detection method based on the information entropy theory combined with the convolutional neural network as claimed in claim 1, wherein: the fourth step, the data with the entropy value smaller than the threshold value is directly output by a convolutional neural network to be classified, and the data with the entropy value larger than the threshold value is subjected to learning by a random forest model and then is classified in the specific process as follows;
(1) The convolutional neural network outputs a data category smaller than a threshold value, and the random forest further learns and divides data higher than the threshold value;
(2) Counting the uncertain sample rate in secondary classification;
(3) And combining and outputting the primary classification result and the secondary learning classification result.
6. The intrusion detection method based on the information entropy theory combined with the convolutional neural network as claimed in claim 1, wherein: the specific process of transmitting the test set after data preprocessing into a trained network model to obtain a classification result is as follows;
(1) Performing parameter tuning on the intrusion detection model by a training set, and obtaining an optimal intrusion detection model when the secondary classification uncertain sample rate reaches the lowest value through adjusting the threshold value;
(2) And inputting a test set to test the intrusion detection model to obtain a final classification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211094816.6A CN115659323A (en) | 2022-09-05 | 2022-09-05 | Intrusion detection method based on information entropy theory and convolution neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211094816.6A CN115659323A (en) | 2022-09-05 | 2022-09-05 | Intrusion detection method based on information entropy theory and convolution neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115659323A true CN115659323A (en) | 2023-01-31 |
Family
ID=84984031
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211094816.6A Pending CN115659323A (en) | 2022-09-05 | 2022-09-05 | Intrusion detection method based on information entropy theory and convolution neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115659323A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116599779A (en) * | 2023-07-19 | 2023-08-15 | 中国电信股份有限公司江西分公司 | IPv6 cloud conversion method for improving network security performance |
-
2022
- 2022-09-05 CN CN202211094816.6A patent/CN115659323A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116599779A (en) * | 2023-07-19 | 2023-08-15 | 中国电信股份有限公司江西分公司 | IPv6 cloud conversion method for improving network security performance |
| CN116599779B (en) * | 2023-07-19 | 2023-10-27 | 中国电信股份有限公司江西分公司 | IPv6 cloud conversion method for improving network security performance |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110598851A (en) | Time series data abnormity detection method fusing LSTM and GAN | |
| CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
| CN107579846B (en) | Cloud computing fault data detection method and system | |
| CN111598179B (en) | Power monitoring system user abnormal behavior analysis method, storage medium and equipment | |
| CN111507504A (en) | Adaboost integrated learning power grid fault diagnosis system and method based on data resampling | |
| CN115952067A (en) | Database operation abnormal behavior detection method and readable storage medium | |
| CN115659323A (en) | Intrusion detection method based on information entropy theory and convolution neural network | |
| CN116226103A (en) | Method for detecting government data quality based on FPGrow algorithm | |
| CN113780432B (en) | Intelligent detection method for operation and maintenance abnormity of network information system based on reinforcement learning | |
| CN114487129A (en) | Flexible material damage identification method based on acoustic emission technology | |
| CN114037001A (en) | Mechanical pump small sample fault diagnosis method based on WGAN-GP-C and metric learning | |
| CN112422546A (en) | Network anomaly detection method based on variable neighborhood algorithm and fuzzy clustering | |
| CN116151799A (en) | BP neural network-based distribution line multi-working-condition fault rate rapid assessment method | |
| CN111031042A (en) | Network anomaly detection method based on improved D-S evidence theory | |
| CN114826764B (en) | Edge computing network attack recognition method and system based on ensemble learning | |
| CN115189939A (en) | HMM model-based power grid network intrusion detection method and system | |
| CN113609480B (en) | Multipath learning intrusion detection method based on large-scale network flow | |
| CN115700558A (en) | Abnormal flow detection method based on space-time feature analysis | |
| CN112788063B (en) | LDoS attack detection method in SDN based on RF-GMM | |
| CN113657520A (en) | Intrusion detection method based on deep confidence network and long-time and short-time memory network | |
| CN118041692B (en) | Network security testing method and system based on intrusion detection technology | |
| CN115996135B (en) | Industrial Internet malicious behavior real-time detection method based on feature combination optimization | |
| CN113422766B (en) | Network system security risk assessment method under DDoS attack | |
| CN117079736B (en) | Gas concentration prediction method and system for intelligent gas sensing | |
| CN116842467A (en) | Network flow anomaly detection and classification method based on bidirectional gating convolutional neural network |
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 |