An auto-learning approach for network intrusion detection

In this paper, we propose a novel intrusion detection technique with a fully automatic attack signatures generation capability. The proposed approach exploits a honeypot traffic data analysis to build an attack scenarios database, used to detect potential intrusions. Furthermore, for an effective and efficient intrusion detection mechanism, we introduce several new or adapted algorithms for signature generation, signature comparison, etc. Finally, we use DARPA'99 and UNSW-NB15 traffic to evaluate the proposed approach. The results indicate that the generated attack signatures are of high quality with low rates of false negatives and false positives.