Abstract
Nowadays, Intrusion Detection Systems (IDSs) are becoming more and more effective since they can benefit from the flexibility offered by Machine Learning (ML) techniques. In this work we investigate the potentiality of the Weightless Neural Networks (WNNs) as a classification method of network attacks. Traditionally, WNNs have been exploited in the image classification field and are implemented through the WiSARD algorithm. Interestingly, our analysis reveals that, applied to the IDS realm, WNNs offer surprising results in terms of performance/time complexity trade-off with respect to other ML-based techniques. The experimental assessment is carried on by considering one of the most updated datasets (CIC-IDS) in the field of the intrusion detection, where two exemplary attacks to be detected are considered: Distributed Denial of Service (DDoS) and PortScan.