Self-Driving Cars May Be Easier To Hack In The 5G Era: Expert

Waymo Google IO AM AH 10 New logo

The fifth generation of mobile networks is likely to make self-driving vehicles more vulnerable than they currently are, according to Martin Hron, a Senior Researcher at cybersecurity giant Avast. In an interview with Android Headlines, Mr. Hron acknowledged that any immediate effects 5G solutions might have on the autonomous driving field are currently hard to predict but made an educated guess that things are simply going to get worse once such technologies are deployed on a large scale. “It’s hard to tell now, but it is highly likely that 5G extends the attack surface by simply allowing more parts and systems of self-driving cars to be connected to the outside world,” the industry veteran said.

The good guys are still ahead of the bad guys

How the cybersecurity space evolves until then remains to be seen and new technologies could certainly alleviate many of today’s concerns stemming from the incoming wireless revolution and how it pertains to the emerging field of driverless transportation solutions. And while more connected components unavoidably leads to less secure technologies, things aren’t so grim in the autonomous car segment; “We are lucky that automotive industry — especially ‘smart cars’ and self-driving cars — is the only industry where we believe the good guys are ahead of bad guys as there are more studies, documented vulnerabilities and proof-of-concepts from security researchers than real, in-the-wild attacks,” Mr. Hron believes.


Now it’s up to the automotive industry to maintain that advantage and stay ahead of the black-hat hacking community by building on the existing studies and attack prototypes, ultimately bettering the security mechanism implemented into connected vehicles, self-driving or not, according to the Avast official. Ensuring transportation methods of the future are as secure as possible from cyberattacks cannot be an afterthought, Mr. Hron warns, elaborating that such protections must be part of the product design process from the very start. Some movement on that front is already happening in the industry, with the popular but largely insecure CAN bus system now being on the verge of being replaced by a new set of internal car bus standards.

The dawn of artificial intelligence isn’t necessarily good or bad

Even today’s connected vehicles are starting to take advantage of AI solutions and the incoming arrival of autonomous vehicles will see AI eventually become an integral part of everyday transportation. AI-driven software isn’t necessarily more or less secure than traditional programs, though its increased complexity could certainly allow for new attack vectors, Mr. Hron explains. E.g. something like computer vision-based perception of a driverless vehicle can be fooled into stopping a car by specifically crafted visuals, though no such attacks have yet been proven to pose actual real-world risks. Still, the fact that AI-based technologies — self-driving or not — operate more like a system or network of computers instead of a single entity consequently makes the attack surfaces they allow for larger, which in turn makes the process of hunting for weak links more likely to succeed.


Self-driving companies must do more

The arrival of 5G, Internet of Things solutions, and everything else the next wireless revolution will enable requires automakers such as former Google unit Waymo to do more in order to ensure their vehicles of the future are as secure as possible, the same Avast official warned in the past. What “more” actually entails is a multifaceted answer, though the industry should at the very least be looking to limit the scope of online communications self-driving vehicles perform so as to reduce the number of possible attack vectors while this particular branch of cybersecurity is still in its infancy, as many experts previously suggested.