Duplicating keys via distant digital images

Beware of flashing your keys in public.

Computer scientists at the University of California at San Diego have developed software that can make a duplicate of a key from just a distant photo of it using technology available to almost anyone.

Referred to as Sneakey, the system is capable of "teleduplication--extracting a key's complete and precise bitting code at a distance via optical decoding and then cutting precise duplicates," according to Sneakey's Web site.

Part of the project's mission is to make people realize that traditional keys are not really as safe as they might think. Relatively modest technology is now capable of the imaging and computer vision algorithms necessary to duplicate an image precisely, according to the group.

To illustrate the point, they photographed a set of keys they casually placed on the table at a cafe from about 195 feet away using a telephoto lens. From that image (shown), they were able to extract enough data to duplicate the keys on the ring perfectly.

The group was able to duplicate keys from a set photographed at about 195 feet away.

(Credit: University of California at San Diego)

It gets worse. The group's software was also capable of extracting enough visual data to make a duplicate key from an image taken by a cell phone camera.

Not only that, but the keys photographed do not even have to be in profile. Sneakey's software can determine a key's bitting code--its series of unique cuts--from nearly any angle.

Stefan Savage, the computer science professor at UC San Diego's Jacobs School of Engineering who led the project, presented his group's work Thursday at the ACM Conference on Communications and Computer Security in Alexandria, Va.

"There are experts who have been able to copy keys by hand from high-resolution photographs for some time. However, we argue that the threat has turned a corner--cheap image sensors have made digital cameras pervasive and basic computer vision techniques can automatically extract a key's information without requiring any expertise," Savage said in a statement.

While the group is not planning to publicly release the code, it inferred in the project statement that anyone with a basic competence in MatLab, a technical computing language and environment from MathWorks, would be able to duplicate its efforts.