About DBIS   | Story archive   | Contact DBIS  | DBIS home

Fake Photos

Computer Scientists Develop Technique for Spotting Digital Forgeries

September 1, 2004

New algorithms spot discrepancies in digital photos. For example, when two images are merged, one or both of the pictures must be shrunk or enlarged to fit together. The new algorithms detect statistical irregularities in the resized photos as well as other signs of tampering.

What is a pixel, and how can it be manipulated?

Science behind the news is funded by a generous grant from the NSF

The word pixel comes from "picture element," and it is the basic unit of color on a computer display or in a computer-generated image. When digital technology is used to capture, store, modify or view images, they must first be converted into numbers: the 1s and 0s found in binary data streams. The images are then put back together in a grid of small squares. These tiny squares are the pixels, and they are the building blocks of all computer text, graphics and images.

A digital camera works just like an ordinary camera, but it does not require chemical development for the pictures. Instead, the images are stored on disks in the computer memory chip inside the camera. Images are recorded by something called a charged-coupled device (CCD): a collection of thousands of light-sensitive cells. When each cell is struck by light, it emits an electrical signal, which can then be converted into a digital data stream.

Before the digital age, it was fairly easy to detect altered photographs. Thanks to the widespread availability of powerful image processing and editing computer software, digital images are easy to manipulate. Any part of the image can be altered, pixel by pixel. Images can be enlarged or enhanced; facial features can be changed to someone's appearance in the image; and brightness can be changed. You can add or remove things from an image with no obvious signs of tampering. As a result, digital forgery has become much more common.

The most common type of digital forgery is called the "copy-move attack," where part of an image is copied and pasted onto another image. There are still clues to the tampering, however. For example, let's say that an image is 10 pixels by 10 pixels, and it is then stretched to 10 pixels by 20 pixels using image-editing software. The image's original pixels will be assigned to every other slot in the new image. This leaves a bunch of "blank" pixels, so the software fills in the gaps based on the surrounding pixels.

What are the signs that a digital image may have been altered?

  • Objects are not in correct perspective
  • Indicators of time, such as clocks and shadows, are inconsistent
  • Shiny surfaces don't show correct angles of reflection
  • There are unexpected discontinuities in the background

The American Mathematical Society contributed the information in the accompanying TV report.


Video help

Latest stories

  • A Satellite Named Violet and a Student Named Amanda
  • Behind the Scenes with the K-Team
  • Deep Space Discoveries
  • Dogs Fighting Cancer
  • Earthquake! What's Your Risk

More information on this story

Martha J. Heil
mheil@aip.org
American Institute of Physics
Tel: 301-209-3088


© 2011 American Institute of Physics