Abstract:

This paper concerns about the problems about processing large\|scale and high dimension network datasets in intrusion detection. The typical feature selection algorithm Fisher was used in network security datasets, in order to reduce the dimension of features. Knearest neighbor algorithm was
used as the classify algorithm, to evaluate the detection rate. A general intrusion detection framework based on feature selection was presented and realized. Experiments  show it has a satisfying  detection accuracy with less features and a good feasibility.

Key words: intrusion detection, Fisher feature selection, Knearest neighbor algorithm