Automated Risk Analysis of Software-defined Systems/Robots
Software-defined systems require a new approach to risk analysis. These systems are characterized by features such as frequent SW updates, AI components, and the use of digital twins, and must be analyzed for reliability and dependability. With each SW update, the behavior of the system can change, thus the risk analysis must be performed in an automated way before each SW update.
The automated generation of advanced hybrid risk models and model-to-model transformation methods are crucial for this purpose. The integration and synchronization of the risk analysis module with the digital twin shall be investigated. AI-based methods are used to extract the required inputs for the risk models from the digital twin. A combination of formal methods and fault injection will be investigated to assess the risk of AI-based SW components. The developed risk models will be able to adapt to the behavior of mutable systems, in contrast to classical risk models.