Skip to Main Content
For the alarm flooding problem, a hierarchical alarm processing model is studied to filter, reduce and correlate alarms. In filtering, false alarms are eliminated with repository. In reduction, a reduction algorithm is designed to remove the duplicate alarms in real time. In correlation, a frequent episodes algorithm is implemented on training data to help clustering-based correlation algorithm find the intrusion patterns. Through the above processing, the false and invalid alarms are eliminated, which eases the networks system and administrator's burden. Meanwhile, intrusion patterns can be found and alarm prediction can be reported. Experimental results show the model is effective.