Nlp-Based Payload Feature Extraction for Enhanced Network Traffic Anomaly Detection

This paper presents a novel approach to network traffic anomaly detection, focusing on feature extraction from both payload and packet-level data. The proposed method utilizes semantic features derived from payload content, combined with statistical and protocol-specific characteristics, to create a multidimensional representation of network traffic. A Random Forest classifier is employed to perform classification tasks, distinguishing between normal and anomalous traffic. Experiments are conducted on the CICIDS-2017 dataset, which includes diverse attack scenarios and protocols, providing a realistic simulation of real-world traffic. The results demonstrate the method's ability to integrate diverse features and improve anomaly detection accuracy. The results demonstrate that payload semantic information is a critical factor in improving the detection of network anomalies.