Every day, news reports detail the arrests of child predators and their horrific impact on innocent victims. Child sexual and physical exploitation is exploding online, and law enforcement officials need new tools to combat it. With an estimated one million predators and only 6,000 specialized law enforcement officials’ worldwide dedicated to fighting child exploitation, agents are outnumbered
“The data is staggering,” said DHS Science and Technology Directorate (S&T) Program Manager Patricia Wolfhope. “The number of reports received through the National Center for Missing and Exploited Children (NCMEC) cyber tip line has grown steadily each year, from 223,374 to 326,310 to 415,650 in 2010, 2011, and 2012, respectively. NCMEC’s analysis indicates the number of images being collected and traded by offenders worldwide continues to expand exponentially, and these images include graphic and violent abuse featuring young children, including infants.”1
Watch the HSI Child Exploitation Investigation Forensics Video
Large amounts of explicit child exploitation material are traded on what is known as the darknet, where predators use encryption, anonymization, and other techniques to make it almost impossible to trace back to the originator. At any given time there are more than 300 darknet boards with more than 500,000 members whose sole purpose is facilitating the exchange of child exploitation material. Further, the same illicit imagery is duplicated, manipulated, and shared among child abuse collectors.
“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.”2
DHS S&T and DHS Homeland Security Investigations (HSI), Cyber Crimes Center (C3), Child Exploitation Investigations Unit are collaborating on automated recognition programs that can help comb through the seized materials. These research and development programs will greatly reduce the amount of time it takes forensic analysts to locate crime scenes, identify and rescue the children, and identify their perpetrators. S&T and HSI C3 are designing, developing, testing, and integrating new face recognition algorithms that will allow agents to sift through massive amounts of digital data much faster and efficiently than current manual processes. Agents could process materials faster by eliminating materials already seized in other cases. S&T expects these technologies to help find investigative leads that would otherwise be missed due to the sheer bulk of data and human limitations. These new innovative technical solutions will increase the number of children recognized and therefore saved from a life of abuse.
To automate a forensic analyst’s work, S&T initiated a project called CHEXIA – Child Exploitation Image Analytics. CHEXIA designs, develops, tests, and integrates new face detection and recognition algorithms. It uses computer algorithms to identify faces in seized material.
S&T’s CHEXIA program has two main parts:
- Face Recognition Algorithms (AFR) algorithm evaluation.
S&T funded the National Institute of Standards and Technology (NIST) to host an industry challenge to assess the capability of face recognition algorithms to correctly detect and recognize children's faces appearing in seized child exploitation imagery. This first required an operational dataset of child exploitation imagery be annotated with ‘ground truth’ information (that is, the key to the test). NIST ran the submitted algorithms against this ground-truth data to determine their performance. The results are available to federal, state, local, and tribal agencies by request. - Integration of Face Recognition Algorithms (AFR) into media forensics platforms.
For this segment of CHEXIA, AFR is being integrated into the existing forensic tools, which are free for all law enforcement agencies worldwide. The true power of AFR is the ability to cluster images with the same face, drastically reducing the amount of time an analyst must spend analyzing massive amounts of data. This allows the analyst to quickly find images of a particular face in question and view all images containing that face for clues. To ensure interoperability, an application programming interface, called JanICE, is being developed to allow any AFR software conforming to the interface to be plugged into the tool.
In May 2017, S&T transitioned the first phase of CHEXIA to U.S. Immigration and Customs Enforcement (ICE) HSI. The next and last phase of the CHEXIA effort will be to continue evaluating face algorithms against child exploitation digital imagery to determine the most effective algorithm for law enforcement and integrate that algorithm. Thus far, the Intelligence Advanced Research Projects Agency Janus algorithms are out performing all other algorithms against this very difficult data set. The CHEXIA Program will be completed in December 2018.
In addition to technological advancements in the fight against child exploitation, in 2018 S&T will also begin applying the social sciences to address behavioral aspects of the problem. First, S&T will develop a research roadmap. Through extensive stakeholder interviews, needs and gaps in research pertaining to child exploitation will be identified. This information will inform and serve as the foundation for future research undertakings.
When it comes to the unconscionable and hideous crime of child sexual and physical exploitation, S&T is developing innovative technologies to help level the playing field for law enforcement agents to fight against perpetrators and save innocent children from a life of torment from their abuse.
For more about the Digital Forensic Program, contact Patricia Wolfhope at first.responder@hq.dhs.gov.
1 “Brief for the National Center for Missing and Exploited Children as Amicus Curiae in Support of Respondent Amy Unknown”. (http://www.missingkids.com)
2 “Child Sexual Abuse: What Parents Should Know,” American Psychological Association. (http://www.apa.org/pi/families/resources/child-sexual-abuse.aspx) (February 19, 2014)