Fighting against Human Trafficking with TrafficVis

Human trafficking (HT) is a devastating problem that affects countless lives globally. Despite the efforts of law enforcement and domain experts, it remains one of the most difficult crimes to detect due to the sheer volume of ads and the presence of legitimate sex workers, spam ads, and other modus operandi (M.O.s) that make it hard to differentiate between real HT cases and false alarms.

TRAFFICVIS

To help tackle this problem, researchers have developed TRAFFICVIS, an interactive visualization tool that leverages advertisement metadata to quickly identify and label suspicious clusters of connected ads as potential evidence of human trafficking. The tool is designed to provide domain experts and law enforcement with a way to visually inspect meta-clusters and label their likelihood of being a particular M.O.

The core of TRAFFICVIS is based on a technique called meta-clustering, which builds on the state-of-theart HT detection algorithm, InfoShield. It groups micro-clusters into larger meta-clusters based on extracted metadata such as images, emails, phone numbers, and social media accounts. This technique helps identify connections between micro-clusters that would otherwise be missed, leading to more accurate HT detection.

Time Savings and Features

TRAFFICVIS provides a vast improvement over manual labeling of ads, resulting in a median speedup of 36x. It takes one person 8 hours to manually label 500 ads, while it takes TRAFFICVIS just 5 minutes to label the same meta-cluster. The tool offers a range of features to help an analyst (e.g., law enforcement agent or HT domain expert) investigate meta-clusters, including basic statistics, micro-clusters, metadata usage, geographical spread, ad text, and the ability to label the meta-cluster on a scale of 1 to 5 for each possible M.O. The labeling panel helps save time and streamlines the process, making it easier for domain experts to work efficiently.

Conclusion

The three major contributions of TRAFFICVIS are an interactive application for HT detection and labeling, meta-clustering to connect ads under the same organized activity, and a significant time-saving speedup over manual labeling. The researchers behind the tool are keen to continue its development, including a formal user study to evaluate its efficacy and usability, and potential collaborations with Marinus Analytics. These further developments would enable researchers to continually develop and evaluate algorithms for M.O. detection, allowing law enforcement to quickly find meta-clusters of ads that represent real HT cases.

In conclusion, TRAFFICVIS is a powerful tool that offers a new approach to detecting and combating human trafficking. Its meta-clustering technique and interactive visualization interface provide a much-needed solution to the complex problem of HT detection, making it easier and more efficient for domain experts and law enforcement to identify and label real cases.