Master's thesis: Interpretable CAN Bus Intrusion Forensics via Tensor Networks

Vehicle CAN networks lack built-in authentication and encryption, leaving them exposed to cyberattacks. Many intrusion detection models identify attacks accurately but offer little explanation of their decisions. This thesis at RISE, within the MAGIC project, will explore whether quantum-inspired tensor networks can combine effective CAN intrusion detection with interpretable, reproducible results that support forensic investigations.

The assignment

• Review research on automotive cybersecurity, CAN intrusion detection, tensor networks and quantum-inspired machine learning.
• Examine CAN attack types such as denial of service, spoofing, fuzzing, replay and masquerading.
• Develop representations of CAN traffic using arbitration IDs, payload bytes, message sequences and timing.
• Implement a tensor-network anomaly detector that learns normal communication patterns.
• Investigate how detected anomalies can be traced to particular messages, bytes and communication patterns.
• Evaluate the framework on public automotive cybersecurity datasets and benchmark it against conventional machine learning and deep learning methods.
• Assess detection performance, interpretability, model compactness, reproducibility and forensic usefulness.
• Document the results and recommend directions for deploying interpretable AI in automotive cybersecurity.

Who we are looking for

• A master's student with foundational knowledge of computer science and cybersecurity.
• Skills in machine learning and programming, for example in Python, along with familiarity with machine learning frameworks.

What we offer

• Collaboration with RISE's Computer Science and Industrial Systems Department within the MAGIC project.
• Close cooperation with the Swedish automotive industry.
• Flexibility to carry out part of the work remotely.

Practical details

Full-time thesis lasting six months and worth 30 credits, for one student. Based at RISE Industrial Systems in Sundsvall or Kista, with partial remote work possible. The stated start is by January 2025, which predates the publication date and needs clarification. The dedicated deadline field states 10 November 2026, while the closing application instructions mention 10 November 2024; the dates are inconsistent. Submit a short personal letter, CV, recent academic transcript and a code sample you wrote yourself. Interviews take place as applications arrive. Questions: Nishat Mowla, nishat.mowla@ri.se, 073 051 19 37.