Using Open Source Facial Recognition Software In Digital Forensics

No ratings

Presented at BSidesDC 2013 by

Brian Lockrey

As hard drives get larger and larger, it is not uncommon in digital forensics to be preented evidence from a system that contains thousands, or even tens of thousands of photos and videos. This talk describes an application built using open source tools to assist law enforcement and investigators in creating a list of suspect images using a database of learned "facial datasets" from evidence collected in other cases. Unlike simple image and video hashtags, this technique is effective even if the image or video has been altered from its previous format. Many times images are resized for mobile devices or videos are converted from one file format to another and this technique will still enable the images to be detected. In some cases, child pornorgaphy can be "hidden" inside of a long running 2-hour movie and by using "facial recognition" across every frame in a video file, it is now possible to detect these snippets. The process also works well on some videos files that may have corruption or missing data that can not be viewed or examined using other software. Created as part of a tool to assist in the detection of child pornography, this software also help investigators locate facial images, even if the images are inside of a file without a standard graphics or video file extension. There are also many other applications for "facial recognition" software including network and computer security, criminal, social media and private investigation purposes.