Modern cars contain a lot of nifty electronic gadgets, as well as more than one kilometer of cable wired to all kinds of sensors, processing units, and electronic control units. The cars themselves have become large computers, and as history shows, wherever there is a computer, there is someone trying to attack it. Over the past few years various studies have been conducted on how feasible it would be to attack a car through its onboard network. Most researchers focused on attacks with full physical access to the car, but some also explored external attack vectors.
If attackers have physical access to a car they can, for example, access the Controller Area Network (CAN) or the On-Board Diagnostic (OBD) system, but they can also perform other dangerous actions, such as physically tampering with the brakes or stealing the car. Digitally tampering with a car, on the other hand, might be much more difficult to prove after an accident. Such attacks could potentially be combined with other attacks that allow for a remote code execution and should be taken as a demonstration of payloads.
There are a few ways to get into a car’s system without having physical access to it, for example through tire pressure monitoring systems, traffic message channel (TMC) messages, or GSM and Bluetooth connections. Some manufacturers have started developing smartphone apps that can control some of the car's functionalities, which opens another possible attack vector. There have also been some cases where specially crafted music files on USB drives were able to hijack some of the car’s systems.
Charlie Miller and Chris Valasek, two researchers working on a project for DARPA, explored how far they could go by hacking the Controller Area Network once inside the car. The pre-released video of their presentation for the upcoming DEFCON conference shows that nearly all of the car's functions can be controlled or triggered including, switching off all lights, shutting down the engine, disabling the brakes, some limited steering, sounding the horn, and manipulating the system display. It doesn’t take much imagination to understand that this has the potential to cause serious accidents. Some of these changes could be made permanent and invisible with malicious firmware updates or system changes. Of course, a laptop with a modem in the glove box would work as well, but would not be as stealthy. If an attacker used the same method as the researchers, hopefully you would notice the attacker’s laptop on your backseat and wonder what was going on.
Car manufacturers are aware of these challenges and have been working on improving the security of car networks for years. Remote attack vectors, especially, need to be analyzed and protected against. At Symantec we are also monitoring this research field to help improve it in the future. Miller and Valasek’s research shows that cars can be an interesting target for attackers, but there are currently far bigger automobile-related risks than hackers taking over your car while driving. Personally, I’m more scared of people texting messages while driving and I assume they pose a far bigger risk than hackers when it comes to accidents, for now at least. Safe driving.