Revealing the Social Structure of Child Sexual Abuse Offending using Digital Forensic Artifacts

• In Event: Child Sexual Abuse Research Reduction Network: Showcasing Research on Technology-Facilitated Child Sexual Abuse and Exploitation

Abstract

Gathering high-quality intelligence information about the offenders and victims appearing in child sexual abuse investigations can be a painstaking and labour intensive process. Recent advances in machine learning have led to the development of new AI software tools that have improved both the scale and speed by which relevant digital artefacts can be identified and analysed. In this paper, we demonstrate how said advances can be used to rapidly isolate discrete persons across a large trove of data, and reveal the contexts through which they are linked to others. This is accomplished here by leveraging a custom-built software system developed by the research team to analyse n=519,776 media files sourced from multiple devices/online platforms seized by Australian law enforcement during the course of a child sexual abuse investigation. The software uses a systematic method to locate and match multiple biometric modalities (face and voice), alongside other file attributes (nudity detection, geolocation and camera sensor information). These data are combined to reveal complex victim/offender networks, and modelled to predict who (and which files) should be priority targets of investigations. The implications of this work are discussed - offering insight into the ways this crime type is organised socially, as well as its practical implications for law enforcement.

Authors

• Russell Brewer, University of Adelaide

• Thomas Swearingen, The University of Adelaide

• Bryce Westlake, San Jose State University

• Katie Logos, University of Adelaide