Traffic flow theory-based anomaly detection models for vehicular networks

Although we drive our vehicles as independent mobile units on roads today, future driving will be a collaborative task among multiple entities. Vehicles will be cooperating as a network of vehicles to make driving decisions that will make future transportation safer, more comfortable, and efficient.  For instance, if the road you are driving on is congested ahead, other vehicles in the affected zone will let your vehicle know about the congestion so that you can replan your route to avoid the congestion and reach the destination with minimum delay. Communication of this nature is known as vehicular communication. Vehicular communication (V2X communication) will facilitate information dissemination among vehicles. A vehicle will be able to transmit and receive data from/to other vehicles and roadside units. This mechanism is known as a vehicular ad-hoc network (VANET).  By facilitating collaborative decision making, VANETs will change the way we drive forever.

However, a vehicle must ensure the integrity of data it receives before consuming them. For example, a malfunctioned sensor or an attacker with malicious intent can inject anomalous data into a vehicular network. Unless these anomalies are detected and filtered out, vehicles can make inaccurate decisions that can result in car crashes, deaths, and damages. Therefore, it is vital to secure vehicular networks against these anomalies.

At rCITI, we develop models to detect anomalies in vehicular networks. These models utilise vehicular physics dictated by traffic flow theory combined with machine learning to identify inconsistent and implausible data being transmitted in vehicular networks. The models we develop help vehicles to make better and accurate decisions and ensure that future intelligent transport systems are secured. 

For more information please contact:

Dr Malith Ranaweera Kankanamge
E: malith.ranaweera@unsw.edu.au

Professor Vinayak Dixit
E: v.dixit@unsw.edu.au