Late Fusion for Improving Intrusion Detection in a Network Traffic Dataset
Addisson Salazar, Nancy Vargas, Gonzalo Safont, Luis Vergara · 2021 International Conference on Computational Science and Computational Intelligence (CSCI) · 2021
This paper presents a method for an intrusion detection system based on late fusion of classifiers. The proposed method was tested in the analysis of a network traffic dataset considering up to 14 classes of anomalous traffic. Several data quality issues were solved by preprocessing: missing data, non-numeric data types, and imbalance of the data classes. The high dimension of the data was reduced by a feature selection method. The results demonstrate the capabilities of late fusion to improve classification accuracy and stability of the intrusion traffic detection from the ones obtained by the individual classifiers.