Digital image forensics in a megapixel world
Leading the Eyewitness
In the May 2007 issue of Scientific Computing, I wrote about the topic of “covered-writing,” a method used to communicate secret messages that
is often known by its expensive name of
“steganography.” Specific bits of a digital
image file that have been replaced with
the bits of a secret steganographic payload
permit a covert agent to post top-secret
documents on their Facebook wall by
simply uploading what appear to be cute
images of kittens on any typical Saturday.
After these images have been shared all
over the Web, the desired recipient downloads them and extracts the stealth payload. In these images, the kittens are used
as cover and are not as important as the
secret information. Fastforward to today,
and an average smartphone purchased
at the corner discount store enables the
addition of kittens to any of
your images using an app and
a single tap of the screen.
While not technically steganography, the original image is
William Weaver, Ph.D.
now hidden by its doppelganger cover.
This past April 1, 2014, Google pulled
a prank on its Google+ users by adding
the Auto-Awesome Photobombing feature
that automatically added images of David
Hasselhoff to users’ private digital photographs. While a fun April Fools prank
for Google, it demonstrates how easy it is
for algorithms to take a digital image of
your favorite political opponent and add
images of unhappy kittens to the scene.
With little effort, a popular faked image
of your challenger being nasty to young
felines can effectively sink their candidacy.
Google’s robots could just as easily place a
street-view image of your vehicle into that
of a busy intersection monitored by a red
light camera. Detecting augmented images,
important since the invention of the photograph, is becoming increasingly difficult.
While use of the copy and paste tool in