Statistics Seminar - Bayesian models of Criminal Plots

Dates

Tuesday, November 23, 2021 - 14:00 to 15:00

Speaker: Jim Smith (Warwick University)

Abstract: Statistical models of criminal activities are challenging to implement. They often need to inform real time decision support tools that are able to identify suspicious activities associated with criminal plots as these progress in time. The observation sequences associated with such threats can be very varied and any underlying sample information is often based on textual data such as testimonies and police reports of past incidents. In this talk I will describe how we are designing new types of dynamic hierarchical Bayesian models to inform these support tools. We are currently applying these to three distinct types of plots: terrorist attacks against the general public, the exfiltration of sensitive documents from computer networks and the illicit manufacture of pharmaceutical drugs at an established chemical facility. All the different applications have been part of co-creation projects with policing organisations - two facilitated by the Alan Turing Institute