Intrusion detection systems (IDSs) were developed for decades and have proven their values in detecting a variety of host and network attacks. Generally, intrusion detection can be considered as a classification problem of identifying incoming events into normal or abnormal, where machine learning is one of the most important tools.

Motivated by the recent development of machine learning techniques, this talk will introduce the current trend of applying machine learning into intrusion detection, and discuss the main limitations and open challenges.