A Web User Interface Tool for Metaheuristics-Based Feature Selection Assessment for IDSs
Humberto Mendes, Silvio E. Quincozes, Vagner E. Quincozes · 2022
Intrusion Detection Systems (IDSs) are intended to identify malicious activities in communications networks and computer systems. In this context, Machine Learning algorithms are powerful tools to support the development of robust IDSs. Nevertheless, their robustness depends upon the selection of representative features for building a precise model for each attack profile. However, choosing an efficient feature selection algorithm is challenging. Whereas filter-based feature selection methods are faster, wrapping-based methods are more assertive. An alternative is using metaheuristics to combine both methods. However, the calibration of such algorithms is challenging, especially because of their lack of usability, which limits developers' implementation and comparison of results. In this work, we propose a friendly web tool for the setup and result assessment of metaheuristics-based feature selection techniques. According to our assessment, GRASP-FS-UI reached a score of 75.8% in Usability and 81% in Perceived Usefulness metrics.