security and privacy of machine learning

Data Ethics and Research Methods
Example topics for working towards cw1 Research Report

# 1 Working Title Keyword 1 2 Keyword 2 Keyword 3
(optional)
1. security and privacy of
machine learning
machine learning security privacy
vulnerability
2. attack and defence in
machine learning
machine learning attack defence
vulnerability
3. adversarial machine learning adversarial machine
learning
4. security and privacy of deep
learning
deep learning security privacy
vulnerability
5. security and privacy of
reinforcement learning
reinforcement
learning
security privacy
vulnerability
6. privacy-preserving machine
learning
machine learning privacy preserving
7. federated learning federated learning 3
8. attack and defence in
federated learning
federated learning attack defence
vulnerability
9. algorithmic bias algorithmic bias machine
learning

Notes:
i) The Boolean operator between multiple keywords should be AND, i.e., conjunction (
).
Remember the calculus of conjunction (
) and disjunction () (OR): Term-1 { Term-2
Term-3 } = { Term-1 Term-2 } { Term-1 Term-3 }, which simply means that there are
two searches, i.e., { Term-1 AND Term-2 }, and then { Term-1 AND Term-3 }.
ii) While the example topics above are all workable, obviously it’s not an exhaustive listing. If
you are not sure about putting something into a relevant topic for cw1, you should seek
comments from the module tutor.
1
The numbers of listing are purely for distinguishing rows of items, with no implication for any order of the
topics.
2
Keyword 1 must be a type of machine learning; otherwise it may not be relevant to “secure AI”.
3
Federated learning is a privacy enhancing approach in itself. Thus, it may go alone without other
keywords.