Tommaso Zoppi

With the growing processing power of computing systems and the increasing availability of massive datasets, Machine Learning (ML) algorithms have led to major breakthroughs in many different areas. This applies also to safety-critical systems, ICT systems whose failure may have catastrophic consequences for the health of citizens, environment, and infrastructures. However, components and software that use ML algorithms may fail: this failure may propagate to the incorporating critical systems and may create severe hazards that need to be avoided. As such, companies, researchers and practitioners are investing in methodologies, architectures and algorithms that are suitable to be safely used in critical systems, albeit no general and reliable solution has been yet found. 

This thesis will make the student knowledgeable about basics of ML and safety engineering. Then, the student will get to know the main design patters for critical systems, which should be re-adapted to work with ML algorithms. Then, the student will perfoem an experimental evaluation aimed at understanding if traditional design patterns (redundancy, safety monitors, recovery blocks) could help in designing safe ML components.