Need to protect drones and robot cars against cyber attacks

Robotic vehicles such as Amazon delivery drones or Mars rovers can be hacked more easily than people think, new research from the University of British Columbia suggests.

The researchers, based at the applied sciences faculty of UBC, designed three types of covert attacks on robot vehicles that caused the machines to crash, miss their targets or complete their missions much later than planned.

The attacks required little to no human intervention to succeed on both real and simulated drones and robbers.

“We saw major weaknesses in robot software that could allow attackers to easily disrupt the behavior of many different types of these machines,” said Karthik Pattabiraman, the professor of electrical and computer engineering who oversaw the study. “Particularly disturbing is the fact that none of these attacks could be detected by the most commonly used detection techniques.

Robotic vehicles use special algorithms to stay on track while they are moving, and to signal unusual behavior that can signal an attack. But a certain degree of deviation from the travel plan is usually allowed to take into account external factors such as friction and wind – and it is these deviations that attackers can exploit to throw the vehicles off course.

The UBC team developed an automated process that allows an attacker to quickly learn the permitted deviations of robot vehicles with conventional protection systems. Hackers can then use the information to launch a series of automated attacks that the vehicle cannot detect until it is too late.

“Robot vehicles already play an important role in surveillance, warehouse management and other contexts, and their use will only increase in the future,” said Pritam Dash, a graduate student in electrical and computer engineering at UBC and the lead author of research. . “We need security measures to prevent rogue drones and robbers from causing serious economic, property and even physical damage.

The researchers offer the basis for a couple of such countermeasures – including self-adjusting deviation thresholds – in a recent document describing their findings. They will present their work at the Annual Conference of Computer Security Applications in San Juan, Puerto Rico, next month.