Secure Autonomous Systems

The course “Security of Autonomous Systems” covers the following topics, which are taught through lectures, tutorials, and assignments:

1. Fundamental Material on autonomous systems: The course starts by discussing the theoretical background of typical autonomous systems, including typical architectures and functional blocks of autonomous vehicles and drones with focus on the layered architecture, simulation environments for End-to-End testing of autonomous systems, notable examples of real-world attacks against autonomous vehicles, and the current state of security research in this domain.
2. Threat Modelling: As basis for further discussion, the course then covers important threat- and asset classification methodologies for automotive systems (e.g. EVITA), as well as their application to autonomous systems.
3. Camera Sensor Processing Pipelines: The course then inspects the typically utilized camera sensor technologies and their underlying security context. This topic area includes the typically performed data (pre-)processing steps, the underlying sensor data processing pipeline between sensor and perception layer, typical perception layer building blocks, security implications of the processing pipeline, and the current state of security research in the context of camera-based sensing systems.
4. Lidar Sensor Processing Pipelines: The course then inspects LiDAR sensor processing pipelines as a second type of important sensing technology of autonomous vehicles. Similar to the camera-based topic area, this topic area introduces the typical sensor data (pre-)processing steps, the typically utilized sensor data pipeline between sensor and perception layer, typical perception layer building blocks, as well as the underlying security implications and current offensive and defensive research topics in autonomous LidAR applications.
5. Sensor Fusion, Collaborative Perception and Real-Time Systems: After establishing the theoretical background for both LiDAR- and Camera-based systems, the course then covers typical sensor fusion approaches, in which these two sensing systems are combined. Next to an overview of such sensor fusion / multi-modal sensing approaches, current research directions and security considerations are discussed. Additionally, typical applications of multi-system collaborative perception approaches are discussed, and current research and security implications highlighted. As autonomous systems typically utilize real-time methodologies, the theoretical background, as well as important attacks and defences for such real-time systems are covered.
6. Wired and wireless communication in the context of autonomous vehicles: The course then highlights the security context of typically utilized wired and wireless communication architectures and protocols found in autonomous systems. To give all students, independent of their exact technical background, the possibility of researching such communication systems as typically found in autonomous systems and their environment, the course offers an introduction into the required RF physics and handling of Software Defined Radios (SDR) using open-source toolkits like GNUradio. Similarly, the theoretical background of wired in-vehicle communication is covered (e.g. CAN and LIN), and the security implications of the typically utilized communication systems and protocols, as well as current research trends are highlighted.
7. Reactive and Preventative Security: To highlight the typically proposed defensive measures intended to protect the discussed autonomous system weaknesses, the course covers both reactive (e.g. automotive IDS) and preventative (e.g. testing-based approaches like simulation-based fuzzing) defensive measures. Current trends in research are highlighted, and the current gaps in the proposed countermeasures are critically discussed.