Secure and Trustworthy AI for Autonomous Vehicle Ecosystems University of Southampton in United Kingdom

This PhD project, supervised by Dr Erisa Karafili at the University of Southampton, aims to advance the security and resilience of AI-enabled autonomous vehicle ecosystems. The research will develop AI-enhanced threat-modelling techniques to systematically analyse vulnerabilities in AI-driven sensing and decision systems, model complex adversarial interactions, and design adaptive detection and mitigation strategies. The project is highly relevant to safety-critical domains, with a focus on real-world applications in the automotive industry.

Autonomous vehicles increasingly rely on artificial intelligence for perception, decision-making, and communication, which introduces new cybersecurity challenges. The project will investigate advanced threat-modelling approaches to improve the security and resilience of these systems under adversarial conditions. Key research areas include identifying and characterising attack surfaces in AI models, with particular attention to threats such as data poisoning, sensor manipulation, and model interference. The propagation of such attacks across interconnected vehicle subsystems will also be studied.

A formal, multi-dimensional threat-modelling framework will be developed to represent interactions between physical, cyber, human, and AI elements within the vehicle and its operational environment. The integration of AI-driven threat intelligence will support automated extraction, classification, and prediction of emerging threats. Adaptive security mechanisms leveraging threat-aware reasoning will be explored to detect anomalous behaviour and inform mitigation strategies, enhancing the robustness of onboard AI systems and decision-making processes.

The project offers an industrial placement with a global automotive manufacturer, providing hands-on experience with real-world autonomous vehicle security challenges and exposure to cutting-edge industrial research. The research will be evaluated using realistic autonomous vehicle scenarios and may be transferred to other safety-critical domains.

Applicants must have a UK first-class honours degree or its international equivalent in cyber security, computer science, artificial intelligence, engineering, or a closely related discipline. Essential skills include excellent programming abilities (Python, C++), strong background in cyber security (threat modelling and secure systems), solid understanding of artificial intelligence and machine learning, and a strong interest in autonomous vehicles and safety-critical systems. Experience with adversarial machine learning or security of AI-enabled systems is desirable. The project is open to UK and Horizon Europe students only. English language qualification is required if applicable.

This is a fully funded IDLA EPSRC studentship with an industrial partner, covering tuition fees and an enhanced stipend above the UKRI minimum. It also includes a generous allowance for travel, equipment, and laboratory consumables. The University of Southampton is committed to equality, diversity, and inclusivity, and welcomes applicants from all backgrounds. Flexible working patterns and career breaks are fully considered.

Applications are considered as received and may close before the stated deadline if a suitable candidate is identified. Early application is encouraged. For general queries, contact [email protected]. For an initial conversation about the project, contact Dr Erisa Karafili at [email protected].