Mapping Modern Slavery
A proof-of-concept study using automatic number plate recognition to map and identify modern slavery vehicular activity.
This project sought to determine the feasibility of using Automatic Number Plate Recognition (ANPR) camera data to identify vehicles involved in the perpetration of modern slavery within two police forces in England. The intended aim was to link police intelligence logs relating to modern slavery with ANPR data and to use machine learning to identify patterns in the driving activity of modern slavery perpetrators and, ultimately, to develop a tool that could be deployed by police forces to identify vehicles as potential perpetrators. For several reasons, the project was unsuccessful in achieving this aim. The report describes these activities and the lessons learned for the predictive application of ANPR and for the future of big data policing.
- The location of ANPR cameras is confidential to police organisations. This constrains the ability for non-police organisations to use contextual and geographical data to inform the predictive models.
- Anecdotally, there are patterns in the logistics of modern slavery offences, but it is unclear if ANPR camera coverage and data are suitable to describe these patterns. Although the use of ANPR for the detection of modern slavery offences was deemed not to be feasible within the constraints of this project, ANPR data may be useful for the detection of other offences that require vehicles to travel long distances, such as human trafficking and smuggling.
- The prevalence of vehicles that are involved in modern slavery offences is likely to be too low to support a useful live detection tool in control rooms, as it would likely generate a large number of false positives alerts. This low base rate means that maximising the specificity of the tool will be challenging.
- If a detection tool for modern slavery is to be based on conviction data, the one-year retention period for ANPR data may be too short for a non-police organisation to develop robust predictive models as the number of relevant convictions within this timeframe is likely to be low at the level of individual forces.
- The launch of the National ANPR Service presents an alternative method for the development of an application of this type across a wider geographic area. A national service could facilitate a more refined algorithm, but would require a national database of offences on which to develop and test models. This would require efficient coordination of national data on modern slavery or other vehicle-based offences.
- Many of the challenges faced by the project could be overcome by the development of data science capacity within police forces or at national policing level.
Authors: Iain Brennan, University of Hull, Lancashire Constabulary and West Yorkshire Police – December 2019
Further information – Iain Brennan (I.Brennan@hull.ac.uk)