Research on Intrusion Detection Based on Optimized Active Learning Algorithm in High-Dimensional Massive Data Environment

In order to improve the correct rate of network intrusion detection (NID) and effectively intercept various network intrusions, an optimized active learning algorithm (OAL) is proposed for the current problem of high dimensional and diverse network intrusion data as well as easy variability, which makes it difficult to obtain ideal network intrusion detection results by a single method. The core of OAL is the introduction of semi-supervised learning (SeS) and active learning (AL). The network intrusion data is clustered according to the network intrusion features using SeS technology, and the clustered data is trained using AL to construct a network intrusion detection classifier. The test results show that OAL can significantly improve the correct rate of NID, shorten the detection time and improve the detection efficiency. Moreover, OAL has significantly lower miss detection rate and false detection rate than classical NID methods, which can better ensure network communication and data transmission security.