ford driverless car

Protecting driverless car data from signal-jamming hackers

Image credit: ford

Researchers have developed a way to combat malicious signal jamming which could disrupt communication networks and even lead driverless cars astray, potentially endangering the lives of others.

Driverless vehicles are designed to collect the same kinds of information that a human driver might, such as traffic lights and the behaviour of other cars, in addition to relying on communications from mobile networks and GPS.

However, ‘adversarial’ signals could jeopardise their operation - a problem that needs to be remedied before autonomous technologies become mainstream.

“The ability to transmit data from a source to a destination reliably in the presence of adversarial intervention, such as jamming, is of paramount importance and concern,” said professor Tamer Başar at the University of Illinois at Urbana-Champaign, lead author on the study.

“The prototype introduced in the paper captures scenarios that arise in many application areas, such as remote sensing systems, networked control systems and cyber-physical systems.”

For example, a sensor may collect information over a period of time and transmit the data to a decision centre that must work to accurately process the original data.

The data can become corrupted as it must be encoded prior to the decision centre and decoded afterwards. Time constraints and limited energy resources further complicate matters: a jammer can stop everything by literally jamming the system with a gluttony of noise.

“The sensor, the encoder and the decoder act in unison, toward a common goal, whereas a jammer operates in a way to counteract the actions of the first three,” Başar said.

The researchers grouped the three pieces together, comprising one actor in the system, working to counter the actions of the jammer. By having all three pieces work as one, they simultaneously announce their policies regarding information.

When the sensor, encoder and decoder work together, they commit to their next actions together. They don’t block the jammer entirely, but the jammer doesn’t have the opportunity to interrupt the work and cause a substantial error as the actors communicate back and forth.

Called a ‘Stackelberg feedback solution’, this hierarchal manoeuvre allows the system to commit to processing information based on a set of pre-computable thresholds, which depends on time and the number of transmission opportunities left. The jammer is left out of consideration as the sensor, encoder and decoder decide together what, how and when to process.

“Our goal is to extend the model introduced in the paper to more complex systems, allowing for more general source processes, multiple sensors, multiple channels and sensors that are equipped with an energy harvester that has the potential to replenish the sensor’s used energy based on random availability of such resources, such as solar or wind power,” Başar said.

In 2017, Nvidia announced a custom chip designed for driverless vehicles that includes an AI ‘nanny’ that constantly monitors how the software is interacting with the vehicle and if it sees something awry, it will attempt to either prevent the unwarranted action from taking place or bring the car to a standstill. 

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles