Research On Intrusion Detection Based On Improved Slime Mould Algorithm

In the era of big data,network attacks often occur,which are fatal to individuals,enterprises and countries.As a dynamic and active defense measure,intrusion detection can effectively prevent most of the intrusions.The upsurge of machine learning and artificial intelligence technology brings more possibilities to intrusion detection technology.Intrusion detection method based on machine learning and intelligent optimization algorithm has become a hot topic among scholars.At present,there are some problems in intrusion detection using machine learning,such as imbalanced data categories,classifier parameter selection and so on,which seriously affect the performance of intrusion detection.To solve the above two problems,this paper proposes an improved algorithm based on slime mold algorithm(SMA),Levy flight,weighted extreme learning machine and other technologies.On this basis,two intrusion detection methods are proposed combined with weighted extreme learning machine and support vector machine(SVM).The main research points of this paper includes the following three parts:(1)Traditional slime mold optimization algorithm uses uniform distribution to update the location,which is easy to fall into local optimum.In this paper,an improved slime mold algorithm(LSMA)is designed.The algorithm introduces Levy distribution,generates random number by Levy distribution,and redesigns the location update formula.In the same environment,13 benchmark functions are simulated,and standard deviation,average,median and best result are used as evaluation indexes.The results show that LSMA effectively improves the probability of escaping from local optimum in the search process.In order to further verify the performance of LSMA algorithm,by using the standard deviation as the evaluation index,experiments show that LSMA has faster convergence speed,higher convergence accuracy,and can find the optimal solution faster.(2)Aiming at the defect of SVM's slow parameter determination in intrusion detection,an SVM intrusion detection algorithm based on LSMA is designed.Taking two parameters of SVM as the individual of LSMA,the optimal slime mold position after iteration is the optimal solution of two parameters.After confirming the optimal parameters,the intrusion detection model is established.The experiment shows that compared with other algorithms,the detection rate of the method is higher and the training time of the algorithm is shorter.(3)Aiming at the imbalance of classification in NSL-KDD intrusion detection dataset,an intrusion detection algorithm based on LSMA weighted extreme learning machine(WELM)is designed.The algorithm optimizes the input weights and offsets of WELM by using LSMA's high probability ability to jump out of local optimization and global optimization.The results show that compared with WELM,the proposed method has a certain improvement in recall rate and false alarm rate.