2 min
To advance AI-enabled cybersecurity research, the National Science Foundation (NSF) presented Kemal Akkaya, Ph.D., professor and chair of the Department of Computer Science, with a $600,000 grant through the organization’s Cybersecurity Innovation for Cyberinfrastructure program. Akkaya’s three-year project will explore how large language models (LLMs) can automate packet labeling for intrusion detection systems. “From transportation and healthcare to finance, improving the accuracy of machine learning algorithms used to defend the networks that underpin these sectors’ cyberinfrastructure is critical for protecting them from cyberattacks. Strengthening these defenses helps ensure the reliability and security of the essential services people rely on every day,” said Akkaya. Intrusion detection systems monitor network traffic to identify suspicious or malicious activity. These systems rely on machine learning models trained on large volumes of accurately labeled data. Producing those datasets, however, is time intensive and often requires expert cybersecurity knowledge. As digital systems increasingly power transportation, health care, finance and communication, the volume and sophistication of cyber attacks continue to grow. At the same time, artificial intelligence is reshaping how both attackers and defenders operate. Improving how quickly and accurately security systems can be trained is critical to protecting the infrastructure that supports daily life. Akkaya’s project will investigate how generative AI can help address this challenge. The team will fine tune open-source large language models using network data, threat signatures and expert annotations. Model accuracy will be strengthened through retrieval-augmented refinement, ensemble modeling and human-in-the-loop verification. Labeled datasets will be released in stages to support the development and evaluation of cybersecurity models. Using data from AmLight, an international research and education network operated by Florida International University (FIU), the project includes collaboration with researchers from FIU. The award strengthens VCU’s growing leadership in AI-enabled cybersecurity research and provides hands-on research training for graduate students. Resulting datasets from this work will support machine learning education for undergraduate students.
