Research On Intrusion Detection And Application Based On TWSVM

Intrusion detection is a key technique in modern computer securityprotection system，and it is also the hot field of network security research．Inrecent years， support vector machines has become a research hotspotin intrusion detection field．However，in the application of intrusion detection，the detection accuracy and training time of the traditional support vectormachines are limited when handling with large amounts of data，which can notmeet demand． Twin support vector machines is the improvementof traditional support vector machines， which reduces the computationalcomplexity of the algorithm and reduces the training time，providing a newapproach for the development of intrusion detection．Based on the analysis of the core idea of twin support vector machines，itsalgorithm is designed and implemented on the Matlab developmentplatform．Besides，a model of instruction detection using twin support vectormachines is proposed．After that，by testing the KDDCUP99data set withtraditional support vector machines and twin support vector machines，results analysis is done by experiment． And its feasibility and validity is proved by experiment． Compared with the traditional support vectormachines， twin support vector machines have a great advantage in thedetection accuracy，false alarm rate，and the training time in which accuracy isimproved，the false alarm rate is reduced，and the training time is reduced byhalf or more，providing an effective method for the intrusion detection systemwhen handling with big data，and also providing an effective solution for theintrusion detection of smartphones based on cloud computing with big data．Based on above analysis，a cloud-based security solution is proposed forsmartphones which can use the model of intrusion detection based on twinsupport vector machines proposed above．The security solution means to putthe security detection for smartphones from itself to the cloud．The solutiontakes the limited resources of smartphones into account and put the intrusiondetection work from the terminal to the cloud．The cloud receives data fromsmartphones and detects data，undertaking the main workload and greatlyreducing the consumption of resources of smartphones．Then， KDDCUP99data samples are adopted to do the classification experiments of traditionalsupport vector machines and twin support vector machines．Data samples hasalready collected from the smartphones with system calls， and theexperimental results validate that the method is feasible．...