Evaluation of a Loss-Sensitive Autoencoder - A Deep Learning Model for Malicious Network Traffic Detection

This paper evaluates a deep learning model, an autoencoder, for detecting malicious activity by identifying the loss produced by various types of malicious network traffic. The evaluation was conducted using four types of malicious network traffic: DDoS, Infiltration, Web application attack, and Port Scan. The proposed autoencoder model exhibited promising results and effectively distinguished between the testing benign network traffic and the malicious network traffic.