TU Wien:Security, Privacy and Explainability in Machine Learning/2025-06-16 Exam-2025S

1) Explain Diro2C and its limitations

2) Choose two algorithm from different learning paradigms (e.g. linear regression or random forest). Explain Federated learning for those algorithms, what they have in common and what are their limitations.

3) (a) What is Model stealing? (b) What is the threat model and describe Model stealing attacks

4) What is watermarking and fingerprinting and in what context was it introduced?

5) How can one defend against Evasion attacks?