Aurora Labs’ Germany chief, Rudolf von Stokar, explains how its self-healing software can help secure connected cars with ever-growing lines of software code.
Despite plans to have fewer electronic control units (ECUs) in cars, there are still likely to be many ECUs in today’s and tomorrow’s connected vehicles. And every ECU requires additional lines of code. Israeli software specialist Aurora Labs says a modern premium car features around 100 million lines of code. The company has developed “self-healing” software to help protect the connected car. Aurora Labs’ managing director for Germany tells automotiveIT International how it works.
How should we imagine your self-healing software platform?
Rudolf von Stokar: The self-healing software Aurora Labs has developed can be used for all software systems and ECUs in modern connected cars. It can detect software failures or the malfunctioning of ECUs with the help of a machine-learning algorithm and can drill down to the level of a single line of code. The software, in effect, heals itself by restoring the ECU to a secure state. Over-the-air updates then make sure that all ECUs in a vehicle are always up to date, seamlessly, secure and without any downtime.
How do you address the security issues that come with OTA? And what about the problems arising from dead spots in mobile network coverage?
The Aurora Labs technology has the advantage that OTA updates can also be done when connectivity is impaired. That’s because we compare software versions and only send the new parts. These are very small and can, thus, also be transmitted when the connection is bad. Even an interruption of the connection is no problem, because the update is only installed when the update file has been completely transferred and tested.
Cyber security and software health are becoming increasingly important because the amount of software code in modern connected cars keeps increasing. Software failures can quickly endanger a car and its driver, because “software on wheels” can quickly lead to a lot of “failures on wheels.” On average, every 1,000 lines of code contain 50 errors and, even with standard quality control, up to 15 percent of errors in onboard code remain hidden. The Aurora labs solution continuously looks at all the code and uses machine learning to fix errors.
You opened an office in Munich last year. How is the German auto industry responding to your solution?
The opening of the Munich office was important for us and it has paid off already. Reactions from German auto manufacturers have been positive and we are already working with some leading companies. One concrete example: A customer in Germany had problems with its airbags, which kept failing. Even after several months of investigations, the company couldn’t identify the cause of the problem. Our self-healing software needed a mere two days to identify the problem, which turned out to be a software issue in the airbag controller. Normally, the airbag controller runs a diagnostics test once a day and then calibrates the sensors. This takes about a second, during which time the airbag is unavailable. The problem was that the controller was running its diagnostic tests 800 times a day, which meant that the airbag wasn’t available for 800 seconds a day. We could quickly identify this software error with our machine-learning algorithm and provide a fix through an update.