Security, North America
This could be a major advantage.
See the pretty Honda Civic in the photograph?
It’s not really a Honda Civic. It’s a Russian tank pretending to be an automobile.
See the American spy satellite? A computer analyzes its images, but the computer can be fooled by clever camouflage.
This is why the U.S. Navy wants to understand how to prevent computerized surveillance systems from being suckered into thinking that tanks are passenger cars.
As surveillance drones and satellites deliver a flood of imagery that overwhelms human analysts, computerized classifiers are being used to analyze the pictures and identify objects such as tanks or missiles. But as with any automated technology, the system can be fooled. The Navy wants to figure out how this can be done, and then modify its reconnaissance hardware and software accordingly.
Recommended: North Korea’s Most Lethal Weapon Isn’t Nukes.
Recommended: 5 Worst Guns Ever Made.
The goal of the Navy research project
is to better understand the mechanisms with which one can trick computer vision classifiers in order to anticipate and counter enemy efforts at camouflaging and misguiding our classifiers. For example, a unit may want to confuse enemy surveillance into thinking that a Honda Civic might be a tank or vice versa.
The Navy notes that as computer Artificial Intelligence (AI) improves, “deep neural networks are achieving recognition of familiar objects with near certainty that might otherwise be unrecognizable to human eyes.” But at the same time, minor changes to an object’s appearance can fool computers. Hence, the Navy wants researchers who can deliberately mislead U.S. computers and thus reveal how the trick is done.