Surely one of the greatest fears of modern parents is that their child will fall prey to an online sexual predator. It's estimated that there are over 15 million photographs of child abuse victims in circulation online, and the very nature of the internet makes stemming the spread of such material a difficult and laborious task for criminal investigators. The development of an automated assistance system for image and video evaluation by Fraunhofer researchers is set to make that task a little easier, and a lot quicker. The desCRY software uses complex algorithms to determine if an image or video from a suspect's confiscated storage medium depicts child abuse, in a fraction of the time currently taken to manually trawl through the hundreds of thousands of files often stored on a typical computer's hard drive.
A study by the Rochester Institute of Technology goes some way to validating parental fear of their child being drawn into a web of child pornography - revealing that 11 percent of surveyed 7-9th graders were asked to talk about sexual matters online and that 9 percent accepted an online invitation to meet someone in person. In its second survey of youngsters between 10 and 17 years of age, the National Center for Missing and Exploited Children reported that 4 percent of young Internet users were asked for nude or sexually explicit photographs of themselves. In its 2010 Annual Report, the Internet Watch Foundation reports the number of unique child abuse domains at 1,351 - with the U.S. being responsible for hosting 42 percent of confirmed child sexual abuse URLs.
For the authorities, cracking down on digital child pornography can be a long, drawn-out process which involves tracking down a source, seizing equipment and then manually sifting through digital images and video content on storage media to look for offensive material. Now, a research team led by Dr. Bertram Nickolay of the Fraunhofer Institute for Production Systems and Design Technology IPK - working with investigators from Berlin's State Office of Criminal Investigation - has developed a software solution to help speed up the process, by automatically picking out images of child pornography from large volumes of data.
It's all in the algorithms
Whereas Paraben's Porn Detection Stick uses algorithms to detect and remove pornographic material from a computer, the Development of an integrated software system to fight child pornography project - or desCRY for short - uses face recognition, skin-tone segmentation, contextual and scene analysis, and object recognition algorithms to identify and mark child abuse multimedia content for further investigation at a rate of up to ten images per second.
The software doesn't just analyze and classify images sat amongst holiday snaps (or video sequences spliced into Hollywood movies) that are stored in obvious places like picture and video folders, but will also peek into hidden archives and email attachments - much like anti-virus software looks through all files for malicious or suspicious files.
The desCRY solution has been tailored specifically for use in criminal investigation, and is said to support typical forensic workflows. Users receive continuous progress updates on an intuitive user interface during a drive scan, as well as a real-time tally of the number of detected files. Results are displayed in an image viewer capable of accommodating several hundred photo icons, and can be filtered according to size and type. Suspicious images are singled out by the software for closer examination - with a single mouse click enlarging icons and a second mouse click storing the find as evidence.
A wide variety of search options are available to the investigator, including content-based data sorting and filtering - allowing results to be grouped according to person, object or location. The design also caters for cross-referencing across current and past case databases, and can be scaled up for a multi-user investigation environment.
Coming soon...
desCRY is currently undergoing realistic field tests with a view to adding the detection system to the investigators' toolkit in the near future. Fraunhofer is also currently in discussion with the SAP software group to make the system available to international criminal investigators. Additionally, there is talk of application opportunities in areas like web filtering and parental control solutions for the home or mobile devices.
More details on the project are available (in German only) online.
