Intrusion Detection Model Based On LDR And Improved PSO-BP

With the rapid development of technology of Internet and incremental requirement of security level for people,applications of intrusion detection become more and more popularization.Given variety of attack means and complicated transformation now,traditional intrusion detection can’t achieves promising effects,so it will start to be developed toward novel directions such as AI and Distributed System,etc.Intelligent Intrusion Detection has nice development potential and value of research among them.The paper presents a intrusion detection model which is combined LDR and IPSO-BP on the basis of in-depth research of existing intrusion detection technology,ANN as a machine learning model included category of AI,Laplacian Eigenmap and PSO algorithm.We will list main research work as follows:(1)We do some deep researches on model and algorithm of PSO-BP and experiments of parameters selection to get optimum model adapting for KDD CUP99.(2)We introduce Laplacian Eigenmap to data dimension reduction and use these postprocessing data for simulation experiments by IPSO-BP.(3)We compare detailedly this model on performance parameters with two intrusion detection models including IGA-TS-BP and H-C4.5-NB in same condition.Moreover,we also compare it on detection rate with KS-ISVM in same condition and different public int threads.We learn from experiment conclusions of this paper that LDR-IPSO-BP Intrusion Detection Model has relatively better average detection rate and detection rate of single class,relatively lower FPR and FNR and some generalization ability,compared with models from some proposed mature intrusion detection models.