no image

Found a way to trick the drones. They are not safe!

A certain combination of colors prevents the cameras Autonomous vehicles to recognize objects.Elena Alekseeva

Materials


Unmanned KAMAZ — already got one!

Researchers from the Munich Institute for intelligent systems max Planck discovered a simple way to cheat highly automated vehicles. Special pattern consisting of bright spots of color, makes it impossible to recognize objects, vehicles. There is a total system failure, the unmanned vehicle can leave streaks or slow down unexpectedly. Something that reminds the effect you get when the stroboscopic flash of a certain frequency cause some people seizures. That is, through vision artificially is called a failure of body functions, in this sense, the human brain and the drone have something in common.

This discovery poses a real threat to security in the use of robotic vehicles on public roads.

According to scientists, in just four hours they managed to create a sample color combinations, causing the unmanned vehicle as akin to panic, and it becomes a security risk. The pattern can be easily applied on t-shirts or make stickers on road signs or shopping bags. Hackers can also take advantage of it, warn researchers.

Researchers from the max Planck Institute wear t-shirts with a colored spot, which hinders the recognition of objects with camera systems. From left to right: Andreas Geiger, Anurag Ranjan, Michael Blake and Joel Janai.Researchers from the max Planck Institute wear t-shirts with a colored spot, which hinders the recognition of objects with camera systems. From left to right: Andreas Geiger, Anurag Ranjan, Michael Blake and Joel Janai.

The problem lies in the imperfection of artificial intelligence, which is revealed in the image recognition. The algorithm uses the built-in camera to monitor the environment, such as a road in front of the vehicle and for detecting obstacles. If recognition fails, the machine-robot at best stops, for safety reasons.

The fact that self-learning recognition system images are vulnerable in General and, especially, in the car, not a discovery.

However, long before serial introduction of robotic vehicles, researchers from the max Planck Institute of engineers warn that interference can be easily arranged, and this is just one example.

Of course, scientists and programmers will eventually solve this problem, but so far it remains. Researchers believe that the goal of automakers is to train your system to be resistant to such attacks.

  • What can go hackers in the world of drones, even scary to imagine!

Photo: automobil-industrie.vogel.de

Залишити відповідь