Firewalls are considered the primary component of Intrusion Detection Systems, as they have the capability to prevent suspicious traffic. Nevertheless, the reliance of most firewalls on static detection criteria renders them vulnerable to new attacks. To overcome this limitation, we present a novel machine learning-based firewall detection system in this paper. Specifically, this system employs three algorithms: random forest, KNN, and Naïve Bayes. These algorithms are trained on a dataset consisting of known malicious and benign traffic patterns to identify and classify traffic as malicious or benign. We then evaluate the performance of each algorithm in terms of accuracy, precision, and ROC. We conducted two experiments to evaluate the efficacy of our proposed approach and to determine the algorithm with the best performance. In the first experiment, we used a dataset containing 11 features, while in the second experiment, we increased the number of features to 13. The results of both experiments demonstrate that our proposed approach was able to detect malicious traffic with high accuracy, precision, and ROC. Additionally, our findings indicate that the performance of the machine learning algorithms improved as the number of features in the dataset increased. Our paper provides empirical evidence of the effectiveness of machine learning algorithms in enhancing the performance of firewalls. By utilizing machine learning-based firewalls, organizations can improve their security posture and protect against emerging threats.

Firewall; artificial intelligence; ML; performance