Watch Video: Session by Charlene Deaver-Vazquez of FISMACS.com Title: New Mathematical Models for Forecasting Cyber Attacks Probability theory gives us the ability to quantify risk and forecast events in a predictable way. These methods have been used in epidemiology, seismology, finance, and even space and nuclear safety analysis. Today a new algorithm gives us the ability to model the phenomenon of one event increasing the likelihood of more such events, referred to as "self-excitement". We witness this when a restaurant seats the first customers in the front window because they intuitively know it increases the likelihood of more customers. We also see it when a tweet goes viral. Applying these models to the cyber realm gives us an unprecedented opportunity to peer into the future with the goals of improving our current decisions and developing mitigations before the risk is upon us.

Charlene Deaver-Vazquez

Charlene has worked in the IT field for over 30 years, and as a Subject Matter Expert for over 12. She created the Probabilistic Risk Model for Cyber Framework (P-RMOD4Cyber). She authored the book "Ensure Your Business Success with Risk-Informed Decisions: the easy way to quantify risk". For the last several years she has been providing analytical services for the Nuclear Regulatory Commission.

From Self-Excitement Algorithms to AI-Powered Threat Prediction

The self-excitement models Charlene describes -- where one event increases the probability of subsequent events -- are now embedded in modern SIEM platforms and AI-driven threat intelligence systems. Machine learning models ingest telemetry from endpoints, network flows, and cloud workloads, applying the same Hawkes process mathematics to detect attack chains in real time. When a phishing email lands, the model increases the probability score for credential theft; when credentials are compromised, lateral movement probability spikes. This cascading probability framework maps directly to the self-excitement phenomenon.

Large language models add another dimension: they can ingest unstructured threat intelligence reports, CVE descriptions, and dark web chatter, then correlate that information with an organization's specific attack surface to generate predictive risk assessments. The mathematical foundation has not changed -- probability theory, self-exciting point processes, quantitative risk modeling -- but the computational power and data volume available to apply these models has increased by orders of magnitude. The convergence of AI, mathematical forecasting, and cybersecurity defense is a core topic on the Morpheus Cyber podcast.

1049