作者: Mike Fugate , James R. Gattiker
关键词: One-class classification 、 Data mining 、 Anomaly (natural sciences) 、 Intrusion detection system 、 Artificial intelligence 、 Pattern recognition 、 Computer science 、 Mahalanobis distance 、 Pattern recognition (psychology) 、 Anomaly detection 、 Benchmark (computing) 、 Support vector machine
摘要: This paper describes experiences and results applying Support Vector Machine (SVM) to a Computer Intrusion Detection (CID) dataset. is the second stage of work with this dataset, emphasizing incorporation anomaly detection in modeling prediction cyber-attacks. The SVMmethod for classification used as benchmark method (from previous study [1]), approaches compare so-called "one class" SVMs thresholded Mahalanobis distance define support regions. Results performance methods, investigate joint detection. dataset DARPA/KDD-99 publicly available features from network packets classified into non-attack four attack categories.