<![CDATA[This study aims to evaluate the effectiveness of network security in protecting company data from cyber threats. The methods used in this research include Gaussian Naive Bayes, Bernoulli Naive Bayes, and Multinomial Naive Bayes. The data analyzed are Fortigate firewall logs from the Intrusion Prevention System (IPS), recording threat activities against PT Nusa Network Prakarsa”s network. Data were collected from January to February 2024, encompassing 43,184 entries containing information on Source and Destination IP addresses, Severity levels, ports, origin countries, attacked Services, and types of malware or viruses. The preprocessing steps included Data Cleaning, Label Encoding, and Feature Selection. The results showed that the Gaussian Naive Bayes model with k = 6 provided the best performance with an accuracy of 0.709. The best features used in this model included “Source IP”, “Destination IP”, “Severity”, “Destination Port”, “Service”, and “Name Malware/Virus”. This accuracy is higher than the average cross-validation value of 0.707. Although there is variation in performance across classes, overall, the model successfully provided reasonably good predictions. This research concludes that using Naive Bayes algorithms is effective in detecting and preventing cyber attacks, thereby enhancing company data security.]]>
Copyrights © 2024
