Victoria Selwyn💙

On the World Stage - Advancing Generative AI for Cyber Security @ IEEE DSC, Tokyo 📢 🎉 At the recent 2024 7th IEEE Conference on Dependable and Secure Computing (IEEE DSC - attend.ieee.org/dsc-2024/), hosted by Japan’s National Institute of Information and Communications Technology (NICT), the Fujitsu Research of Europe team was proud to represent our latest cutting-edge research into modern generative AI for cyber security applications. Ofir Manor from Fujitsu Research of Europe – Israel presented a paper from the Advanced Cyber Security (ACS) team as part of the Machine Learning and AI in Cybersecurity session, involving inputs from across Fujitsu Research in Israel, the UK, and Japan. Titled ‘Fine-Tuning Large Language Models for Network Traffic Analysis in Cyber Security’, the paper discusses the use of fine-tuning to assist LLMs in analysing and understanding the technical text (request a copy here - research.europe@fujitsu.com). We showcased this by fine-tuning a GPT model (specifically GPT3.5-Turbo) on network packets that were part of normal (benign) behaviour and those sent as part of 8 distinct cyber security attacks over an HD camera surveillance network. We have shown that the fine-tuned LLM can detect malicious unknown network packets, and in case of malicious behaviour, the fine-tuned LLM can classify the attack type. In our experiments, we have shown that the fine-tuned LLMs performed far better than their non-fine-tuned counterparts. The team is enormously proud of this achievement, representing a major stepping stone in our goal of finding ways the generative AI can assist in cybersecurity tasks. Authors: Ortal Lavi, Ofir Manor, Andrés M. Murillo, Motoyoshi Sekiya, Junichi Suga, Kenji Hikichi, Yuki Unno and Dr. Ayoub MESSOUS #AI #cybersecurity #GenAI #LLM @Fujitsu_Global