Connected vehicles will soon become a reality on our roads, bringing new services, capabilities, technical challenges, and security threats. CARAMEL has developed several anti-hacking solutions for the new generation of vehicles, which will be presented at the IoT Solutions World Congress 2022 in Barcelona to overcome these risks
The project, running from October 2019 to June 2022, applies a proactive approach based on Artificial Intelligence and Machine Learning techniques to detect and prevent potential cybersecurity threats to autonomous and connected vehicles. Visitors to IoTSWC 2022 will have the opportunity to gain a deeper understanding of CARAMEL solutions by taking part in the daily guided tours.
The vehicle part features an architecture based on a secured On-Board Unit (OBU) connected to an additional small computer, the “anti-hacking device”, that can execute security functions.
In the fixed infrastructure part, CARAMEL relays in a Multi-access Edge Computer (MEC) connected to the cloud, where it deploys a backend for monitoring purposes, and a Public Key Infrastructure (PKI) complying with ETSI standards to protect vehicular transmissions, with an additional Certificate Revocation List (CRL) distribution system.
CARAMEL also develops a Vehicular-to-Everything (V2X) radio technology interoperability solution based on a forwarding functionality in the MEC, between the two current mainstream technologies based on IEEE 802.11p and Cellular V2X (C-V2X). Another crucial topic addressed by the project is the acquisition of the correct localization of the vehicle. CARAMEL proposes two solutions for a spoofing attack on the vehicle’s Global Navigation Satellite System (GNSS). One is executed in the anti-hacking device of the car and uses local data, and the other runs in the MEC using a collaborative approach based on V2X information.
CARAMEL also provides new algorithms to deal with the challenge of detecting objects through the vehicle’s LiDAR. The Vision-based Cyberattack Component (VCC) developed in the CARAMEL project explores state-of-the-art methods to protect against vision-based attacks on Connected and Autonomous Vehicles (CAVs). The VCC considers two types of threats: external and internal attacks. External attacks are physical attacks carried out in the environment. They are an indirect attack on the CAVs system, where the attacker modifies environmental attributes such as road signs to alter the vehicle behavior. On the contrary, internal attacks target the internal components of the vehicles, in particular, camera sensors. These attacks require direct access to the camera sensor, aiming to alter captured images.
The VCC was developed to counter such attacks and addresses three major scenarios: physical-based adversarial attacks, image deterioration attacks on camera sensors, and finally, adversarial attacks on camera sensors. Attacks on traffic signs were selected to demonstrate the physical-based adversarial attacks. In contrast, attacks on-camera images using noise and adversarial Deep Learning models were selected to exhibit image deterioration and adversarial attacks. The architecture of the related Deep Learning models, data generation processes, training, evaluation, and finally, implementation of the prototype
CARAMEL pilot demonstrations will occur at the Test Area in Hessen in Germany and at GreenFlux R&D laboratory in the Netherlands, where real-world scenarios will be performed with prototype vehicles and smart chargers under a controlled environment.
Additionally, CARAMEL partners have recently released a series of video demonstrations, each focusing on a different system developed within the project, and conducted an online workshop for Original Equipment Manufacturers (OEM) to discuss the results. As the three-year project moves towards its closing date, CARAMEL results are expected to decisively contribute to European competitiveness in artificial intelligence-based cybersecurity for connected and autonomous vehicles. Video demonstrations are now available on YouTube.