Police forces plan to use artificial intelligence to speed up their investigations

Coordinated by the Idiap Research Institute, the European project Roxanne gathers a large variety of national and international police forces, including Interpol, as well as scientists, large industry and private companies. Its aim is to create a computer programme able to help investigators to link various clues and uncover criminal network activities.

Sometimes, despite a bugged cell phone, due to nicknames and encrypted apps, police officers can’t identify all the members of a criminal network under investigation. Investigator’s experience and flair can help to discover the link between various clues appearing as independent. But in a globalized and hyper connected world, the amount of data to consider is growing too fast. Criminal networks are also crossing borders, making it difficult for police services to work without their foreign colleagues’ help. Considering the discrepancy between the various clues gathered by various people, having a computer programme able to ‘read and understand’ all the data is a real challenge.

Funded by the H2020 European Commission programme in the framework of the “Technologies to enhance the fight against crime and terrorism” call, the Roxanne project’s goal is to boost a collaboration between the actors at the European level to produce a useful tool for police services. The project consortium includes 24 international partners and is planned to last 3 years.

Cold cases and sensitive data

In order to create an ‘artificial flair’, scientists must teach the computer to identify the meaningful links from data sources (audio, video, text) collected by the investigators. To do so, they use machine learning approaches. One of the main challenge is to test it. Due to the sensitive nature of the criminal cases, as well as due to the fact that information can be classified, researchers need to explore novel approaches. For example, they can create fictitious cases or use cold cases shared by police investigators, following strict ethical and legal requirements. Another work option is to install a test version of the programme on the police officer premises so they can run it by themselves and provide a feedback about its efficiency and eventual bugs.

Another challenge lies in bridging various data gathered during an investigation. The aim is to combine strengths of the speech and language technologies, visual analysis and network analysis. The meta-analysis of these elements can provide more accurate information about a suspect, such as the gender, age, or origin of the person. Even if such computer programmes already exist, they are not tailored for specific needs of criminal investigation. Moreover, they often lack transparency about how they work, which doesn’t allow the end-users to properly evaluate the results provided by the computer.

A tool to speed up investigations

The analysis of data related to an investigation casework is often tedious and expensive, involving several human investigators. The use of a specific programme can not only save money, but can also save time to prevent criminal acts, or to help the victims with closure. As a criminal investigation might lead to a prosecution and a trial, there are also ethical and legal considerations. Therefore, the developed tools will predominantly be used by the police services during the analysis phase related to an investigation. The decision making process and remains in human hands. The extraction of an evidence to be used in court is a subsequent step, not directly targeted by the Roxanne project.


More information

Project’s website

Interpol’s news about its involvement

H2020 call