Self-Driving Cars Might Be More Secure Than We Thought
Multiple sensors allow for redundancies that may keep automated cars safer from hacking.
There's a growing concern that autonomous cars will be prime targets for hackers, a fear that is magnified amidst tragic terrorist attacks in which vehicles are used as weapons.
But some security and self-driving experts contend that cars commanded by sensors and software may actually be safer from cyberattacks than vehicles driven by humans. "One interesting thing about fully self-driving cars is they're unintentionally more secure, which is really not what you would expect at all," Craig Smith, a security researcher and car hacker, tells The Guardian.
Smith, whose day job is research director of transportation security at the cybersecurity firm Rapid7, is one of the "white hat" hackers helping automakers secure their vehicles. He's also involved with the Car Hacking Village at Defcon in Las Vegas and author of The Car Hackers Handbook.
So he knows a thing or two about how cars can be compromised. And while having more sensors on a car potentially increases the attack surfaces or entry points for hackers, Smith and others contend that they also provide valuable redundancies and a kind of checks and balances that can thwart hacking attempts.
A Computer Overriding a Human
"The way cars work today is you have a few sensors," Smith tells The Guardian. He used the example of a back-up collision-detection system that can sense objects in a car's path and automatically apply the brakes to avoid a collision. "It's a computer overriding a human."
According to Smith, a hacker may easily manipulate a single sensor and fool the car into doing something it's not supposed to. But self-driving cars use multiple sensors since no single device can do it all.
"Each of the sensors used in autonomous driving solve another part of the sensing challenge," Danny Atsmon, head of self-driving vehicle testing company Cognata, tells The Guardian. Radar sensors are great at detecting objects at great distances but can't discern whether there's, say, a small animal or a paper bag in the road ahead, while a camera has less range but can better determine what the object is.
As a result, "a sensor redundancy and sensor fusion approach" not only allows self-driving cars to make better decisions, but also helps keep vehicles secure from hacking, Cognata says.
"The interesting thing that happens is that each sensor doesn't trust the other," Smith says. If a radar sensor "sees" something ahead that can cause a car to act—such as slamming on the brakes or plowing ahead at full speed—a camera sensor can also weigh in on the decision. It's a human-like reaction, which makes it "harder for a hacker to deal with," Smith says.