• United States



Executive Editor

Comprehensive software security for cars will take years

Mar 30, 20164 mins

Professor who pioneered car software security and broke Viagra spam botnets wins computer science prize

Software security for automobiles is improving but it will take another three or four years until manufacturers can put overarching security architecture in place, says Stefan Savage, winner of the 2015 ACM-Infosys Foundation Award in the Computing Sciences.

“We’re at a point where the industry has to recognize that this is a real issue for them,” says Savage, a professor in the Computer Science and Engineering department at the UC San Diego Jacobs School of Engineering.

+ MORE CAR SECURITY: Car hackers urge you to patch your Chrysler, Ram, Durango, or Jeep +

What’s needed is the capability to update software routinely in order to patch newly found vulnerabilities, similar to the way Microsoft’s Windows Update works, he says.

In addition, a single team needs to oversee all the software that is running on individual components in cars to make sure that they and their interfaces with other devices are secure. “Almost all the vulnerabilities we found were at the interfaces of code written by different parties,” he says.

And often software is written using off the shelf code that hasn’t been stripped down to eliminate attack vectors. For example a device in a car that needs Bluetooth support may contain a Linux Bluetooth stack that has more features than necessary that could represent a vulnerability to the cars’ systems overall.

Without an overarching review of software, vulnerabilities can leak in with after-market parts. One aftermarket CD player, for example, could be updated with a specially formatted CD that enabled the device to override other software systems in the car. “These kind of issues abound,” he says.

Short-term, car makers can improve their patching programs. A bit longer term they can adopt strategies that are common to PC security today but that haven’t necessarily been applied to cars. And longer term, they can produce cars with a whole-vehicle software security scheme, he says.

Self-driving cars pose a different problem. Given the range of capabilities they have that rely on software, it will be difficult to keep them free of bugs and vulnerabilities in traditional ways. Pausing or quarantining software exhibiting anomalies while a vehicle is traveling 70 mph is an obvious risk. Perhaps, he says, these cars will have a limp-home mode to get them off the road safely when issues arise.

Likely car makers will introduce autonomous features like lane changing and parking one at a time to see how well they are received and how reliable they are. If all goes well more and more will be added, leading to self-driving vehicles.

Savage’s other work includes going after Viagra spammers, not by taking down their botnets or blocking their emails but by crippling their financial backend. He analyzed their business models and discovered that their merchant bank accounts used to collect credit card payments were vulnerable. With the cooperation of Visa, Mastercard and their member banks, the team shut down these accounts, making it difficult for them to reap their profits.

+ ALSO: Most ‘hackable’ vehicles are Jeep, Escalade, Infiniti and Prius +

Savage says he hopes the notoriety the ACM-Infosys Foundation prize gives him will help spread the word about an ongoing study to measure empirically the effectiveness of various means of fighting cyberattacks.

The research is measuring how using individual security tools and practices actually affects security. For instance, do the use of antivirus software, patching and steering clear of malicious Web sites actually reduce instances of devices being taken over or credentials being lost? This will help security pros focus their efforts where they will do the most good.

Today, much of common practice is driven by marketing claims of vendors selling security products. “I think you can look in aggregate at what behaviors matter more,” he says.