Centre for Connected & Autonomous Automotive Research (CCAAR) PhD Projects
Dependability Assurance for Autonomous Vehicle Safety
Student: Luis-Pedro Cobos Yelavives
The project aims to find a way to show how dependable vehicle assurance is. During the project a method has been developed that takes the aspects of Classical Safety (active & Passive), SOTIF (Safety of the intended function), Functional Safety, and Cybersecurity. The developed method uses the structure of GSN (Goal Structuring Notation) in combination with ADT (Attack Defence Trees) and challenges every claim, evidence and argument with inductive logic to reduce bias. So far there are 2 examples and a demonstration of the method
- The traffic Signal Recognition System
- AI driving
- Pilot Demonstration.
- Safety and Security Updates of a safety Critical Function like the Airbag
- Test Bench of Updates
- Ecu Analysis
Machine learning generation of attack trees
Student: Kacper Sowka
The ultimate objective of this research is to produce a comprehensive machine learning supported attack tree generation methodology for the automotive cybersecurity domain. A principal aim is to explore how such a procedure could be utilised in practically viable cybersecurity assurance initiatives within the automotive industry, particularly in relation to the recently published ISO/SAE 21434 standard for automotive cybersecurity and UNECE Regulation 155. Of particular interest are methods for the encoding of salient cybersecurity critical information into individual vulnerabilities and the training of a machine learning model which can discern the relationship between two given vulnerabilities or weaknesses.
- Design of generation methodology
- Implementation of methodology with an automotive relevant example
- Sourcing of dataset to train machine learning models
- Comprehensive validation strategy determining the performance of the proposed methodology
Model Based Security Testing
Student: Esma Kalir
The project seeks to augment the existing design methodology for automotive embedded systems with model-based security testing. This will increase the confidence that automotive systems are secure by design and that security properties are realised in the implementation.