By Kyle Haefner, Ph.D., CableLabs; Chad Schwenke, CableLabs
Network traffic analysis plays a critical role in cybersecurity. Artificial Intelligence (AI) and Machine learning (ML) models are increasingly deployed to classify and identify malicious traffic. However, these models are susceptible to adversarial attacks where slight alterations in the data can cause the model to misclassify attacks. Adversarial AI in network traffic involves crafting malicious network traffic that appears benign to ML-based intrusion detection systems (IDS) and traffic classifiers. This paper explores how attackers can manipulate network traffic data to bypass detection and achieve their goals. We discuss techniques for generating adversarial network traffic and the challenges defenders face in mitigating these attacks.