Neural network techniques for host anomaly intrusion detection using fixed pattern transformation

摘要: The weak foundation of the computing environment caused information leakage and hacking to be uncontrollable. Therefore, dynamic control security threats real-time reaction identical or similar types accidents after intrusion are considered important. As one solutions solve problem, studies on detection systems actively being conducted. To improve anomaly system using calls, this study focuses techniques neural networks fuzzy membership function Soundex algorithm which is designed change feature selection variable length data into a fixed learning pattern. That is, by changing sequential call behavior pattern algorithm, conducted back-propagation function. proposed method N-gram technique applied for calls Sendmail UNM demonstrate its performance.