Sophos AI team to present at CAMLIS

The Conference on Applied Machine Learning in Information Security (CAMLIS) is an annual event scheduled to take place in Arlington, Virginia this year on October 19th and 20th. This conference serves as a platform for exploring the convergence of Machine Learning and cybersecurity. The gathering facilitates collaboration among professionals in Applied Machine Learning from academia, research institutions, and the information security industry.

At Sophos X-Ops we strongly advocate for open research and the exchange of expertise, and we recognize the profound importance of the ML-Sec realm in effectively safeguarding our customers’ systems. In line with this commitment, CAMLIS serves as a crucial component of our Sophos AI team’s strategy, facilitating our connectivity to cutting-edge research in this domain. In addition, it grants us an invaluable platform to publicly showcase and share our own research, encouraging collaborative enhancements from fellow professionals.

CAMLIS also offers a remarkable platform for researchers and engineers at Sophos AI to engage with counterparts from other security companies and academic research laboratories. This conference facilitates networking and invigorating discussions on the future of AI in the realm of cybersecurity. The provision of technical talks and deliberations creates an excellent avenue for us to present our work, fostering collaboration with likeminded researchers grappling with similar problems and potential challenges.

The presentations and poster sessions at CAMLIS provide a comprehensive overview of the cutting-edge advancements in the ML-Sec domain. This year, there are notable sessions specifically dedicated to Threat Hunting and Infrastructure for ML-Sec, demonstrating the latest developments in these areas. Additionally, a few sessions will delve into the realm of Large Language models, the transformer-based Neural Networks that have garnered significant attention in the Machine Learning community.

Three of our researchers from Sophos AI will be delivering notable talks at the conference:

• Tamas Voros will present our research on addressing the Web Content Filtering challenge through the distillation of Large Language Models, a topic we’ve covered in a report earlier this year.
• Adarsh Kyadige will discuss a carefully curated series of three benchmarks developed by our team to assess the capabilities of Large Language Models in the realm of Cybersecurity.
• Ben Gelman, another Sophos AI researcher, will present a poster showcasing our development of a tool that seamlessly generates scam campaigns automatically, through the strategic utilization of text and image generation models—a topic Gelman and Younghoo Lee presented on at DEF CON’s AI Village in August.

We will be publishing reports on both the benchmarking efforts and the AI-driven scam campaign research soon.