Research on the Application of CRFs Based on Feature Sets in Network Intrusion Detection

This thesis puts forward a method of CRFs (Conditional Random Fields) based on feature sets in network intrusion detection. This method takes advantages of the CRFs models which can stitch to sequence data marking and add random attributes. It uses varied connection information and its relativity in network connection information data sequence as well as the feature sets relativity to attack detection and discovery of abnormal phenomenon. It uses KDD Cup 1999 data sets as experimental data and comes to a conclusion that our proposed method is practicable, reliable and efficient.