Unmasking Human Trafficking Risk in Commercial Sex Supply Chains with Machine Learning.

This article focuses on leveraging machine learning and massive deep web data to characterize recruitment and trafficking risk in online commercial sex supply chains. Using a novel framework that combines natural language processing, active learning, and network analysis, the study identifies deceptive recruitment posts—such as nonsex job offers linked to sex sales by the same entity—as proxies for human trafficking risk. The approach reveals geographic patterns where recruitment often occurs in economically constrained suburban areas, while sex sales concentrate in large urban centers, and uncovers likely trafficking routes connecting recruitment and sales locations. Synthetic experiments demonstrate that incorporating network-based uncertainty metrics in active learning improves identification of high-risk trafficking pathways and diverse recruitment tactics. These findings offer law enforcement and policymakers data-driven insights to better coordinate countertrafficking efforts and tailor interventions across regions, while acknowledging limitations related to data coverage and evolving criminal behaviors.