Law enforcement test AI tool AviaTor to fight child sexual abuse
The AviaTor project began in January 2019 with project partners from ZiuZ Forensics, Web IQ, the German Research Centre for Artificial Intelligence, the National Police of the Netherlands, Belgian Federal Police and INHOPE.
Law Enforcement Agencies (LEAs) across the globe receive reports of potential Child Sexual Abuse Material (CSAM) from the National Center for Missing and Exploited Children (NCMEC). Companies in the United States like Facebook and Google are required by law to proactively scan their systems for potential CSAM. This results in a huge quantity of NCMEC reports. In 2018, NCMEC’s CyberTipline, INHOPE's U.S. hotline, received more than 18 million reports of potential CSAM from US based companies. The large increase of material every year requires the use of prioritisation tools to swiftly identify children in real danger. The AviaTor project was conceptualised with this purpose in mind.
AviaTor, which stands for Augmented Visual Intelligence and Targeted Online Research, was developed from a research project into a working prototype in just 12 months. By using expert insight gained from extensive interviews with the Dutch and Belgian police, it was possible to establish how to best process NCMEC referrals.
“AviaTor is a tool that is being built for the future.”
Yves Goethals, Judicial Commissioner with Belgian Federal Police.
In addition, European LEAs were invited to provide input on the current process, comment on its strengths & weaknesses, and opportunities through online interviews in exchange for early access to the tool being developed. In total, input was gathered from 16 LEAs. AviaTor has been developed with scalability in mind, with the aim to create a broadly applicable and accepted tool. In total input was gathered from 16 LEAs. It is clear that reports of CSAM – from industry, the public and NGOs – are labour-intensive to process and AviaTor aims to develop automation and intelligence to reduce the time spent by LEAs assessing and prioritising reports. The National Police of the Netherlands have been using the first version of the AviaTor tool since December 2019.
“More effective police time and an increase in capacity means more cases handled, more victims rescued, and more offenders caught.”
Jaap van Oss, Head of Team National CSE Unit, Dutch National Police
Continuously evolving, the tool is being refined through a process of development, integration and delivery. Through this testing period - which is crucial to ensuring the value and success of the project going forward - the Dutch and Belgian police are testing the prototype tool with real reports. However, some countries will not be able to do this and it is therefore essential to have the support of Europol and INTERPOL in training the classifier with a database of baseline CSAM.
Some LEAs are currently having to use temporary solutions to classify and prioritise cases, which is a short term-fix. AviaTor is the long-term solution for law enforcement with its combination of artificial intelligence techniques to investigate the content of imagery, and the capability to carry out targeted online research to enrich the data with open source information in order to create more actionable intelligence.
The AviaTor Project is funded by the European Union’s Internal Security Fund – Police. If you are a LEA representative and are interested in learning more about this project and would like to engage with our partners, please contact us by sending an email to firstname.lastname@example.org.
Photo by INHOPE, European Commission, ZiuZ, WebIQ, AviaTor
The AviaTor project was presented at the Europol EMPACT.'