Situation Elements Extraction Based On Fuzzy Rough Set And Combination Classifier

With the rapid development of modern computer technology and network,it has penetrated into all walks of life.With the advent of the information age,people are not only enjoying the convenience and quickness brought by the network,but also being threatened and perplexed by various network security problems.In recent years,network attacks are becoming more and more complex and covert,which makes the traditional network security products and technologies based on passive defense difficult to deal with the current network security problems,and the network security situation awareness technology based on active defense can better solve such problems.Since the network security situation awareness technology was proposed at the end of the 20 th century,it has been widely used in various fields of network security.Generalized network security situation awareness technology is divided into three processes: situation element extraction,situation understanding and situation prediction.Situation element extraction is the first and the most critical step in the whole process,and its extraction quality will directly affect the accuracy of situation understanding and prediction.Many scholars have put forward many methods to extract situation elements based on various theories.These methods have some advantages in some aspects,but they also have some limitations.In view of the shortcomings of current situation elements extraction methods,this thesis makes an in-depth study on the network security situation elements extraction algorithm,and proposes a situation elements extraction model based on fuzzy rough set and combined classifier,which is used to improve the accuracy of situation elements acquisition,so as to provide a better data basis for situation understanding and prediction.In this thesis,the theory of fuzzy rough set and combination classifier is introduced into the process of network security situation elements extraction.Using the theory of fuzzy rough set,the attribute reduction of data is realized without reducing the ability of data classification,which reduces the complexity of data.Using the theory of combination classifier and particle swarm optimization algorithm,a situation elements extraction framework is built,which can extract situation elements more accurately.The main research work is as follows:Firstly,this thesis proposes a method to measure the similarity of conditional attributes,which is used to calculate the similarity matrix of conditional attributes.Then,the direct clustering algorithm is used to cluster the similar conditional attributes according to the threshold value,and the proposed maximum similarity criterion is used to select the most representative attribute from each similar attribute set to replace the attribute set,so as to achieve the purpose of secondary reduction of conditional attributes.Secondly,in order to reduce the probability of samples being misclassified,ensure the maximum membership of samples belonging to the real class,and reduce the influence of noise samples,this thesis improves the upper and lower approximation algorithm of fuzzy sets,and a method based on k-order distance weighted average is proposed to calculate the upper and lower approximation of close proximity domain.The calculation method is applied to fuzzy rough set to reduce the secondary condition attributes heuristically,so as to obtain the final condition attribute set.Thirdly,this thesis use experiments to compare 15 commonly used classification algorithms,and then according to some metrics of classification algorithm,4 classification algorithms with good performance are selected to construct a combined classifier.In order to make the combined classification results more effective fusion,a BP neural network fusion algorithm based on particle swarm optimization is proposed.The improved particle swarm optimization algorithm is used for fusion training of the BP neural network,which speeds up the convergence speed of the model and makes the situation elements extracted more accurate.Finally,the framework is implemented in code and tested on NSL-KDD,and compared with a variety of attribute reduction algorithms and situation element extraction algorithms.From the experimental results,it can be seen that the network security situation element extraction framework proposed in this thesis can effectively shorten the extraction time of situation elements and improve the accuracy of situation element acquisition under the premise of ensuring the ability of data classification,thus proving the effectiveness and feasibility of the situation elements extraction framework proposed in this thesis.