Abstract

As telecommunication and information networks, which connect modern society through computers, smartphones, and other electronic devices, grow, so do security issues and cybercrime. The seriousness of cyberattacks in the context of network security has been highlighted by this rise in cybercrime. Simultaneously, machine learning has been thoroughly investigated for intrusion classification, with an emphasis on improving the accuracy of classifiers and the effectiveness of data mining models. This paper investigates the derivation of intrusion attack classification rules with Rough Set Theory (RST), a rule-based decision-making approach. To determine the efficacy of four different algorithms, including the Genetic Algorithm, experiments were carried out utilizing datasets that were made publicly available. When the Genetic Algorithm was used for rule construction in RST classification, the results were the best when compared to other rule generating techniques. Furthermore, this approach's capacity to precisely forecast all forms of attacks was shown when it was applied to a dataset of intrusion attacks that was made available to the public. Additionally, this methodology gives security professionals and developers useful information ahead of time, allowing them to take proactive measures when needed.