Logo IMG


Digital Forensics

Modern crime often leaves an electronic trail. Finding and preserving that evidence requires careful methods as well as technical skill

Simson L. Garfinkel

Image Integrity

2013-09GarfinkelF7.jpgClick to Enlarge ImageEven when photos and video can be recovered from a subject’s computer or cell phone, another question to consider is whether the imagery is real. Photographs were doctored long before the advent of Photoshop. For example, in the era of Soviet Russia, after he was purged by Stalin, Bolshevik official Avel Enukidze was carefully removed from official photographs through a series of skillful manipulations of light and negatives in a Kremlin darkroom. Computer animation now takes such manipulation to a completely new level, with synthesized scenes that are visually indistinguishable from recorded reality.

Image processing advances have made it possible to find some kinds of artifacts that indicate tampering or wholesale synthesis. Light reflections, highlights, and shadows also can be closely examined to reveal that different objects in a single “photograph” actually were assembled from images that were in slightly different physical environments.

In one dramatic demonstration in 2009, computer scientist Hany Farid of Dartmouth College showed that a single “group picture” had been created by pasting in people from different photos because the reflection of the room lights on each person’s eyes were inconsistent with were they were standing in the frame.

At the leading edge of digital forensics research are systems that attempt to assist an analyst’s reasoning—to find evidence automatically that is out of the ordinary, strange, or inconsistent. Such details can indicate that there is a deeper, hidden story. Inconsistencies can also indicate that evidence has been tampered with or entirely falsified. Ultimately, such automated reasoning systems are likely the only way that today’s analysts will be able to keep up with the vast quantities and increasing diversity of data in the coming years. Progress in this area remains tricky, however. Some developments have been made on systems that can find timestamp inconsistencies (a file can’t be deleted before it is created), but such rules are invariably complicated by the messiness of the real world (for example, daylight savings time).

comments powered by Disqus


Subscribe to American Scientist