Investigating crimes against children, specifically sexual solicitations, are complicated because not all offenders are contact-driven, meaning they want to meet the minor for sex in the physical world; instead, some offenders are fantasy-driven, in that they are more interested in cybersex and role-play. In addition, the sheer volume of cases involving the online sexual solicitation of minors makes it difficult for law enforcement to determine whether an offender is contact-driven vs. fantasy-driven. However, research shows that there are language-based differences between minors and contact-driven offenders vs. fantasy driven-offenders. Thus, we developed the Chat Analysis Triage Tool (CATT), a forensically sound investigative tool that, based on natural language processing methods, analyzes and compares chats between minors and contact-driven vs. non-contract driven offenders. Using an SVM classifier, we were successful in differentiating the classes based on character trigrams. In a matter of seconds, the existing algorithms provide an identification of an offender's risk level based on the likelihood of contact offending as inferred from the model, which assists law enforcement in their ability to triage and prioritize cases involving the sexual solicitation of minors.
Keywords: Chat, machine learning; Child sexual exploitation; Digital forensics; Internet crimes against children; Linguistic inquiry and word count; Sexual solicitations.
Copyright © 2019 Elsevier B.V. All rights reserved.