Digital Evidence Analysis: Automatic Human Image Detection

On this page, find:

Challenges for Law Enforcement in Detecting Human Images

Most computers have thousands of images stored on the hard drive and investigators must manually search through these files when looking for evidence. Current tools are not specific enough to automatically search through files, and commercially available search programs are neither integrated with law enforcement tools nor customizable for law enforcement use.

This process is very time consuming and error-prone. Often, the images are being searched for evidence of child pornography — one of the most common forms of computer crime investigations. In fact, due in part to the sheer magnitude of the time a child pornography investigation takes and the seriousness of the offense, many law enforcement officers spend what amounts to a majority of their time on these investigations, frequently at the expense of performing other types of investigations.

Human Image Detection Software to Aid Investigations

To assist investigators, NIJ is funding the development of a software program that could identify images based on certain shapes, tones and contours. The tool will be disseminated at no cost to investigators through the Cyber Cop portal Exit Notice.

By not requiring investigators to manually examine every image file on a device, an automated tool could save hours of manpower. Once the program has completed its analysis, an officer would then go back and review the flagged images.

Computer scientists at the University of Rhode Island, supported by a NIJ grant, are working on an automated human image detection tool that could:

  • Mark files with high concentrations of skin tones.
  • Assign probability of digital enhancement or altering, important for image authentication.
  • Be customized to help investigators specify their searches.
  • Be integrated into existing forensic software suites.
  • Provide an easy-to-use interface.
  • Determine if image of a person was computer-generated or natural.

Researchers are seeking ways to improve the tool's ability to locate images of children so that child pornography and other sexually exploitive photographs of minors can be identified.

Date Created: November 5, 2010