A weighted intrusion detection model of dynamic selection

摘要: In view of the difficulty existing intrusion detection methods in dealing with new forms, large scale, and high concealment network behaviors, this paper presents a weighted model dynamic selection (WIDMoDS) based on data features. The aim is to customize models for sets different types, sizes structures. First, according features, single classifiers are clustered using hierarchical clustering algorithm evaluation indicators, then, by means accuracy classifiers, addition, data-classifier applicable indicators (DCAI) performances used calculating weights subjective objective, then combined weight ranks. Finally, custom generated Weight-voting (W-voting) algorithm. Our experiments show that can optimize number reduce problem redundant or insufficient ensemble process. A combining classifier characteristics dataset attributes improve detection.