FULL PDF
Abstract
The proliferation with encrypted network traffic, traditional packet inspection mechanisms fall short in detecting anomalies and intrusions. This paper explores the integration of deep packet inspection (DPI) and unsupervised machine learning methods for detecting network anomalies, even when payloads are encrypted. The study highlights key challenges in feature extraction, proposes a model combining statistical flow features with unsupervised clustering, and validates it on real-world datasets. Results show over 90% detection accuracy without reliance on decryption, making the model promising for future scalable intrusion detection systems.
View more >>