Ant tree miner amyntas: automatic, cost-based feature selection for intrusion detection

Abstract

Intrusion Detection Systems (IDSs) analyse network traffic to identify suspicious patterns which indicate the intention to compromise the system. Traditional detection methods are still the norm for commercial products promoting a rigid, manual, and static detection platform. This paper focuses on recent advances in machine learning by implementing the Ant Tree Miner Amyntas (ATMa) classifier within intrusion detection. The proposed ATMa use Ant Colony Optimisation and a cost-based evaluation function to automatically select features from a data set before inducing Decision Trees (DTs) that classify network data.