The team considered a prototypical intelligent transportation system with a control law that is specifically designed to avoid collisions. They then experimentally demonstrate that, nevertheless, an attack on a position sensor can result in collisions between vehicles. This is a consequence of the feeding of malicious sensor measurements to the controller and the collision avoidance module built into the system.
This is an instance of the broader concern of cybersecurity vulnerabilities opened up by the increasing integration of critical physical infrastructures with the cyber system. Using a solution based on “dynamic watermarking” of signals to detect and stop such attacks on cyber-physical systems. They show how dynamic watermarking can handle nonlinearities arising in vehicular models. The researchers then experimentally demonstrate that employing this nonlinear extension indeed restores the property of collision freedom even in the presence of attacks.
Recently there has been great interest in automated as well as semi-automated transportation systems involving various driver assists. These advanced systems rely on sensors to provide state information and situational awareness to the control logic governing the vehicle. However, this has also increased the vulnerability of these advanced transportation systems to cyber attacks.
In fact, there have been demonstrated cyber attacks on automobiles in the recent past , , where two hackers have remotely subverted an automobile, taking control of its steering and braking units. This ultimately led to the automobile manufacturer recalling over a million cars to patch the identified vulnerabilities. Several other similar reports – point to the need for cybersecurity of automated transportation systems.
In this paper the team demonstrate show the technique of “dynamic watermarking” can be employed to secure an automated transportation system against arbitrary attacks on its sensors.