Blog

We’re not waiting until April Fools for misbehavior detection

As we approach April 1st, we need to be prepared that not everything we read is true. We let everyone decide how much fun they find in silly fake news these days, but such indulgence in the field of road safety cannot be allowed.

Published March 19, 2021

Misbehavior detection is when a message is sent between vehicles and then flagged as false. Misbehavior can endanger lives, therefore both malicious and unintentionally wrong V2X messages must be filtered out on all 365 days of the year.

There’s an app for that!

Well, not on your phone, I’m not referring to the decade old commercial. It’s an app in our vehicle-to-everything stack, and it’s called misbehavior detection.

An early example of misbehavior detection is basically all about checking the values in consecutive Cooperative Awareness Messages (CAM) to expose a liar. For a starter you have to know that CAM messages contain a lot of information about a vehicle: position, heading, speed, acceleration and the vehicle’s type. The message says a car is reversing towards us at 200 km/h? Obviously nonsense. Traveling with a normal 50 km/h, but the GPS coordinates are always the same? Hardly believable. How about a 23 meters long motorcycle? Not even Batman’s bike in the movie reached that size. 

So these were the easier cases.

At the last Car2Car week event Andras Varadi, head of Commsignia research and Daniel Magyar research intern gave a presentation about much more complex issues involving multiple sensors and V2X units.

Infrastructure as a vigilante

In the first case the infrastructure helps drivers assess the situation. The information sharing is done with Cooperative Perception Message (CPM) and, as the name implies, these messages contain data about objects detected by various sensors and cameras. In our case the V2X roadside unit (RSU) at the intersection is connected to a camera. We simulate a case when its input is not verified and as a consequence it is possible that it sends a corrupted CPM message to vehicle B. The message says that vehicle A occupies the intersection.

But vehicle B’s own on-board sensors detect free space at the phantom car’s alleged location. Our misbehavior detection algorithm doesn’t just make a decision based on the vehicle’s own sensor, but also reports the RSU to the Misbehavior Authority. 

Denouncing a V2X device is an integral part of V2X safety, it prevents additional future false messages from being accepted by anyone in the traffic. However, since this process can take time, Commsinia software also helps maintain safety right away. What’s more, vehicle B shares its own view of the situation – that there is nobody in the intersection – using CPMs to contradict the roadside unit.

In the second simulated case, vehicle A plays the culprit. It sends a corrupted CAM message with wrong speed and location data, claiming that it’s fully stopped in the middle of the intersection. The RSU works correctly, recognizes that the vehicle is already away and there’s a drivable free space at the intersection. So the RSU sends out a CPM message containing freespace information.

caption

Vehicle B receives both the wrong CAM and the right CPM messages, and the misbehavior detection algorithm running on the vehicle’s V2X onboard unit checks both data.

Only Commsignia’s special algorithm allows vehicle B to recognize the misbehavior from the pairing anomaly, and warn the driver of a potential danger.