CCC - Data Science Analysis Methods for Law Enforcement

Within the Center for Computational Criminology (CCC) at Blekinge Institute of Technology (BTH), research and collaboration have been conducted together with law enforcement agencies, mainly the police, since 2012. The work was funded by the Swedish Agency for Economic and Regional Growth between 2012-2014, by Region Blekinge in 2015 and by Vinnova in 2016. The purpose of the work within CCC is to investigate, develop and evaluate automatic methods to simplify criminal investigation work within law enforcement agencies.

Within the CCC, active work is being carried out in collaboration with industrial players as well as the public sector to move from research to innovation in the classic triple-helix spirit. In this work, BTH has been given a central role in linking national stakeholders who in various ways work with IT-based methods for crime investigation and prevention, something that was previously lacking in Sweden.

Project leader for the project is Associate Professor Martin Boldt. The project's contact person within the Swedish police is Johan Sundqvist at the Police Development Center South.

Overview of the project

Based on the annual National Safety Survey conducted by the Swedish National Council for Crime Prevention (BRÅ), it is estimated that about two million crimes are committed annually in Sweden and of these, just under one and a half million are reported to the police. Of the reported crimes, around 88% are so-called volume crimes, which consist of, for example, various types of robbery, vandalism, assault and drunk driving. The so-called volume crimes are of a simpler nature and affect a large proportion of the population each year. The police have a low clear-up rate for several of the categories of crime included in volume crime. The aim of the CCC work is to investigate whether computer science methods can assist the police in their crime coordination of crimes (i.e. linking crimes to series committed by the same offenders). The project is divided into the following four sub-projects involving five senior researchers, two PhD students and a number of police officers:

1. Structured data collection from crime scenes.
2. Automated methods for crime coordination.
3. Automated methods for shoeprint matching.
4. Establishing a center for police IT-based methodological research

Project summary

The sub-project on structured data collection includes a process that ensures that the same type of information is collected in a structured way from different crime scenes. By coding the collected information always in the same way, e.g. the modus operandi (MO) of the offender is always recorded in the same way, it is later possible to perform automatic comparisons of crimes. In this way, it is possible for computer programs to automatically search the collected crime scene information in order to identify similarities between different crimes that may indicate that they are part of a common series. Computer algorithms can thus identify possible links between crimes, which police crime coordinators will then have to analyze and evaluate based on their experience, and finally either reject or accept alleged series. This can hopefully lead to the police being able to more effectively link similar crimes to new series committed by common perpetrators. By being more able to convict offenders for series of residential burglaries rather than single crimes, the hope is to increase the clearance rate in this crime category. This will also address organized crime, which is responsible for a reasonable proportion of residential burglaries. This work also involves the development of methods that can increase the efficiency of the registration of crimes in RAR and DurTvå. These methods can eventually shorten the time it takes for a police patrol/locus/technician to record data from a crime scene, while increasing the quality of the data based on automatic checks in the digital cross-referenced crime scene forms used.

It is important to emphasize that these automated methods of crime scene coordination are in no way intended to replace the professional expertise of the police. Rather, they will be tools that can indicate where likely links between individual crimes exist, which in turn the human crime coordinator can analyze and evaluate. These indications of links between crimes increase the possibility of linking important information fragments that are scattered between several crime scenes, which may differ greatly in terms of both geographical and temporal distribution. The more information fragments from different crime scenes that are linked together in relevant series, the better, because this increases the understanding of how offenders behave within the crime category, but also because it increases the likelihood of finding sufficient circumstantial evidence and evidence to achieve a clearance. In addition, joint investigation of multiple crimes allows for a more efficient use of law enforcement resources. Both the automated crime coordination and shoeprint matching methods are to be used as purely internal police tools in the form of selection tools that allow the police to select from their large amounts of information. They are not intended to be used as a basis for e.g. District Court proceedings etc.

The collected crime scene information can also be used to evaluate the effectiveness of crime prevention measures. We are also investigating the possibility of assessing the risk of burglary using the collected data together with general crime statistics in the residential area. That is, specific characteristics of a dwelling together with general crime statistics for the area where the dwelling is located. In this work, we also collaborate with a couple of the largest insurance companies in Sweden.

CCC also involves collaboration with the National Forensic Center (NFC), formerly SKL, in Linköping with the aim of developing methods for automatic comparison of photographs of shoe prints. Shoe prints are one of the more common traces in certain types of crime, but the type of trace has so far required manual comparisons. This in turn has resulted in the comparisons being difficult to scale up to a common national level. Therefore, the projects are studying how such comparisons could be automated in all, or parts, of the process.