The Role of Weibull Distribution for Anomaly Detection in Internet Traffic
Sumayya Zafar, Asad Arfeen · IETE Journal of Research · 2025
In an era where safeguarding network infrastructures against evolving cyber threats is paramount, anomaly detection in Internet traffic emerges as a critical defense. This study introduces a novel approach, leveraging the Weibull distribution, to enhance anomaly detection capabilities. This research’s key aspects include the comparison of seven Weibull distribution estimators, specifically for Internet traffic data, identifying the rank correlation, L-moment, and maximum likelihood methods as exceptionally accurate, based on bias and root mean square error (RMSE) metrics. Using precise parameter estimates, the study successfully detected anomalies within the CICIDS 2017 dataset, achieving accuracy rates of 83.4% (L-moment), 64.4% (Maximum Likelihood), and 81.5% (Rank correlation). Further validation of the model on the CSE-CIC-IDS 2018 dataset – containing attack profiles absent in CICIDS 2017 – demonstrated its robustness and generalizability, with the accuracy rates of 96.6% (L-Moment), 82.3% (Maximum Likelihood), and 93.8% (Rank Correlation). Moreover, with this approach offers a computationally efficient solution with O (1) complexity, crucial for real-time detection. These results position the Weibull distribution as a promising tool for strengthening anomaly detection mechanisms, offering a computationally efficient solution for enhancing network security.