Dutch police used deep learning model to predict threats to life

The Netherlands Forensic Institute (NFI) developed software to help Dutch police filter life-threatening messages sent by suspected criminals using the encrypted EncroChat phone network. After placing a "software implant" on an EncroChat server in Roubaix, French investigators began collecting live data from phones on 1 April 2020, which they then shared with Dutch police through a secure computer link. With the infiltration of the network leading to the interception of at least 25 million messages, Dutch police wanted a way of predicting which messages contained serious threats to life so they could take action. To do this, the NFI's forensic big data analysis (FBDA) team modified a computer model it had previously developed in late 2019 to scan for drug-related messages sent between suspected criminals in large volumes of communications data, as part of a research and development project. EncroChat, which had 60,000 users worldwide and around 9,000 users in the UK (see distribution map below), was used by organised crime groups for drug dealing, money laundering and plotting to kill rival criminals.