@article{MFFE4A1A4,
title = "A Survey on Deep Learning-Based Image Approaches for Malicious Network Traffic Detection",
journal = "The Journal of Korean Institute of Communications and Information Sciences",
year = "2025",
issn = "1226-4717",
doi = "10.7840/kics.2025.50.11.1739",
author = "Young-joo Park, Sun Woo Lee",
keywords = "Malware Traffic Detection, Malicious Traffic Detection, Image-Based Classification",
abstract = "As cyberattacks grow more sophisticated, detecting malicious behavior through network traffic has become increasingly critical. Recently, image-based detection approaches that convert raw network traffic into grayscale or RGB images for deep learning classification have gained attention. This paper provides a structured survey on such methods, categorizing existing research by model architecture (e.g., CNN, CNN-LSTM, GAN, GNN), image processing techniques, target environments such as IoT, and evaluation settings. We also examine open-set recognition (OSR) approaches aimed at identifying previously unseen attacks. By analyzing preprocessing strategies, image encoding methods, and OSR applications, this study outlines key limitations of current research and discusses future directions, including lightweight modeling for IoT, real-time performance, protocol independence, and enhanced OSR techniques."
}