Cybersecurity presents a range of statistical challenges arising from the scale, complexity, and dynamic nature of modern digital systems. Data are often high-volume, heterogeneous, and arrive in real time, requiring methods that are both computationally efficient and robust to noise. Many cybersecurity applications involve limited or no labelled data, making unsupervised and semi-supervised methods essential. Analysts must often work with evolving threats and incomplete information, requiring models that adapt over time and provide interpretable outputs to support decision-making. These challenges place statistics at the core of developing effective, scalable, and explainable solutions for securing digital infrastructure. This webinar will explore some of these challenges, demonstrating how statistical methods can address key problems in cybersecurity.

Daniyar Ghani (Imperial College London) will present statistical models for analysing command logs from honeypots using topic modelling and self-exciting point processes. Lekha Patel (Senior Statistician, Sandia National Laboratories) will discuss real-time anomaly and disturbance detection enabled by cyber-physical digital twins. Leigh Shlomovich (Head of Research, Alpha Level Security) will focus on statistical methods for classifying alerts, particularly in unsupervised settings, and on developing meaningful, explainable approaches to alert prioritisation.

This event is jointly supported by the Emerging Applications Section and Business & Industrial Statistics Section Group.