DHS Science and Technology Directorate developing algorithms to identify child exploitation victims

The Science and Technology Directorate (S&T) of the Department of Homeland Security (DHS) has initiated a project to design, develop, test, and integrate new face detection and recognition algorithms for use in identifying victims of child pornography circulating on the dark web.

The project is called Child Exploitation Image Analytics (CHEXIA), and it is intended to automate the work of forensic analysts, as part of a collaboration between DHS S&T and the Cyber Crimes Center (C3) Child Exploitation Investigations Unit of DHS Homeland Security Investigations.

There are over 300 active darknet boards with over 500,000 members dedicated to child exploitation at any given time, according to the announcement, and S&T Program Manager Patricia Wolfhope says the number of reports received by the National Center for Missing and Exploited Children (NCMEC) cyber tip line has increased steadily each year.

“Locating the perpetrator and victim as quickly as possible is critical,” said Wolfhope. “As it is estimated by the Centers for Disease Control that approximately one in six boys and one in four girls are sexually abused before the age of 18.”

CHEXIA consists of an evaluation of facial recognition algorithms, performed by NIST as an industry challenge, and the integration of the algorithms into media forensics platforms. The Janus algorithms developed by the Intelligence Advanced Research Projects Agency (IARPA) are providing the best results against the data set so far. An API to ensure interoperability is in development.

DHS says S&T will also apply the social sciences to the problem this year. The CHEXIA program is expected to be completed in December, 2018.

Related Posts

Article Topics

 |   |   | 

Comments

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Most Read This Week

Biometrics Research Group

Biometrics White Papers

Biometrics Events

Explaining Biometrics