摘要
The popularization of shared networks and Internet usage demands increases attention on information system security, particularly on intrusion detection. Two data mining methodologies - Artificial Neural Networks (ANNs) and Support Vector Machine (SVM) and two encoding methods - simple frequency-based scheme and tf×idf scheme are used to detect potential system intrusions in this study. Our results show that SVM with tf×idf scheme achieved the best performance, while ANN with simple frequency-based scheme achieved the worst. The data used in experiments are BSM audit data from the DARPA 1998 Intrusion Detection Evaluation Program at MIT's Lincoln Labs. © 2004 Elsevier Ltd. All rights reserved.