KLASIFIKASI SERANGAN SIBER PADA LALU LINTAS JARINGAN KOMPUTER MENGGUNAKAN ALGORITMA K-NEAREST NEIGHBOR DAN GAUSSIAN NAIVE BAYES BERBASIS DATASET CIC-IDS

Abstract

Escalating cyber attack sophistication necessitates highly accurate and responsive Network Intrusion Detection Systems (NIDS). This study examines the classification efficacy of K-Nearest Neighbor (KNN) and Gaussian Naive Bayes (GNB) classifiers in identifying normal and malicious traffic patterns (DDoS, PortScan, and Brute Force) based on the standardized CIC-IDS2017 benchmark dataset. The corpus of 5,000 telemetry flows underwent ANOVA F-value feature filtering selecting the 15 most discriminative attributes alongside Min-Max scaling. Experimental evaluations demonstrate that KNN (k=5) attained exceptional classification performance with 98.40% accuracy, 98.20% precision, and a 98.30% F1-score, significantly outstripping Gaussian Naive Bayes (88.10% accuracy). While KNN incurs higher latency, its robustness in isolating DoS vectors validates its deployment viability within campus perimeter firewalls.