Monday, August 7, 2017

Student Security Team Proves That Stickers on Road Signs Can Dangerously Disrupt Autonomous Vehicles

Simple Vandalism Could Present Challenge to Future Technology
Shipping News Feature
US – WORLDWIDE – Recent reports of major computer hacks and bitcoin ransomware has illustrated the vulnerability of modern technologies and the latest work undertaken by US students will do nothing to boost confidence. The Handy Shipping Guide has covered on multiple occasions developments in autonomous road haulage trucks and other vehicles that do not require a driver in order to make their deliveries. Despite a whole host of development currently being undertaken there is no doubt that there is a long way to go, as a team from several American universities has just demonstrated.

Using simple sticker additions to several types of signs the team was able to fool an autonomous guidance program into thinking that the signs where indicating something else. For example, and rather worryingly, a stop sign that was altered was read by the guidance system to instead be a 45mph speed limit sign. The consequences of such a mistake in the real world would be potentially catastrophic, especially if the vehicle was one of the heavy freight vehicles that are being experimented with.

The team also experimented with creating fake signs to see how well the computers dealt with them (a possible future tactic for hijackers) and, again rather worryingly, found that quite small stickers, the sort that might be placed on a sign as low grade vandalism, could disrupt the systems’ ability to read the sign correctly.

Of course, this experiment only demonstrated how camera systems on autonomous vehicles can be fooled, and most systems are linked to integral GPS and road maps that give the computer additional information on details such as speed limits and road restrictions that they are passing through. However, with visual sensors playing a key part in the autonomous vehicles decision making, the ease that the tests threw them off would indicate that these systems still have a way to go before they will be fit to completely replace humans behind the wheel.