By leaving the steering wheel under the control of an onboard artificial intelligence (AI), designers of autonomous vehicles open the door to a range of vulnerabilities that could compromise both the safety of vehicle occupants and that of people nearby. Gabriela Nicolescu, professor in the Department of Computer Engineering and Software Engineering at Polytechnique Montréal, protects the code of these true computers on wheels.
A car unable to recognize a stop sign at the corner of the street. Another for whom a pedestrian has become invisible. A truck that accelerates at the sight of a red traffic light.
These hypothetical scenarios could become reality if hackers targeted the onboard software code of an autonomous vehicle. Since each company develops its own tools, it is imperative to ensure that they do not contain any vulnerabilities.
This is notably the work of Prof. Gabriela Nicolescu. Specializing in Internet of Things cybersecurity, Prof. Nicolescu collaborates with a major American software-hardware company to ensure the security of its tools intended for autonomous vehicles.
"Autonomous vehicles operate with limited resources, notably in terms of computing power, which forces designers to compress and simplify the code of onboard AI tools as much as possible," explains Prof. Nicolescu. "And by adapting this code, you eventually create vulnerabilities."
Malicious modifications or code injection, data theft, extraction of intellectual property: these are just some of the threats hanging over the onboard AI of autonomous vehicles. Threats that are particularly significant for vehicles' computer vision systems, according to Prof. Nicolescu.
"By attacking the algorithm of this system, hackers can make it so the AI is no longer able to recognize signs or traffic lights, for example," she says. "The consequences could be dramatic."
Competitive reasons are also at play, adds the cybersecurity specialist from Polytechnique Montréal. "Intellectual property issues are significant because an algorithm that requires very little computing power will be very valuable to whoever owns it," she explains.
People flow
As part of another project involving an industrial partner, Prof. Nicolescu's team is working to secure a decision-support tool that supervises in real time the movements made by buses within the same fleet. A tool that also monitors the routes taken by users in a station to help improve the flow of their travel.
"Our role here is less a cybersecurity issue than a safety issue," admits the specialist. "The circuits of the kiosks can be prone to failures, so we make sure to closely monitor each one's energy level, their temperature, etc., to detect problems before they affect the proper functioning of the system."
"These AI tools work thanks to sensors that generate enormous amounts of data," she adds. "The work here mainly consists of optimizing the code to reduce the system's overall energy consumption and detect failures in these embedded systems."
Learn more
Expert profile ofProf. Gabriela Nicolescu >>>
This partner content was produced as part of the En route! project – The career destination in electric and smart transportation.
