Evaluating the effectiveness of machine learning algorithms in detecting malicious traffic on corporate networks

Abstract

The growing complexity of cyber threats imposes significant challenges on organizations that rely on digital infrastructures for their operations. Traditional signature-based intrusion detection methods have proven limited against sophisticated and dynamic attacks, especially those enhanced by automated techniques. In this context, Machine Learning algorithms emerge as a promising alternative for identifying anomalous patterns in network traffic. This article aims to evaluate the effectiveness of Machine Learning algorithms in detecting malicious traffic in corporate networks through a qualitative bibliographic research grounded in scientific literature from the last twenty years. The analysis encompasses comparative studies on the performance of supervised and unsupervised models, evaluation metrics, and practical implementation limitations. Results indicate that models based on supervised learning exhibit high predictive performance; however, they face challenges related to scalability, explainability, and adaptation to zero-day attacks. It is concluded that the effectiveness of these algorithms depends not only on the computational architecture adopted but also on organizational maturity and corporate data governance.