Recently introduced legislation, such as the GDPR in the European Union, includes provisions that require the right to be forgotten. This requirement mandates that organisations take “reasonable steps” to achieve “the erasure of personal data concerning [the individual]”. The unprecedented scale at which Machine Learning (ML), however, is being applied on personal data motivates us to examine how this right to be forgotten can be efficiently implemented for ML systems.
Recent research has demonstrated several machine unlearning approaches [1-3] to implement the “right to be forgotten”, whereas it is not yet clear how good these approaches are at erasing necessitates knowledge of individual training points contributing to parameter updates, from ML models.
This PhD project aims to study and rigorously evaluate the recent advance in machine unlearning. The candidate will also gain an in-depth understanding of the state-of-the-art membership inference attacks [4] on ML, which will be used to evaluate the effectiveness of these machine unlearning approaches in achieving “right to be forgotten”. The outcomes of the evaluate will guide the candidate towards building a new sustainable machine unlearning approach for resource-constrained ML systems
Notes on Machine Unlearning can be found at [5].
Academic Qualifications
A first degree (at least a 2.1) ideally in Computer Science or Data Science with a good fundamental knowledge of machine learning.
English Language requirement
IELTS score must be at least 6.5 (with not less than 6.0 in each of the four components). Other equivalent qualifications will be accepted. Full details of the University’s policy are available online.
Essential attributes
- Experience of fundamental machine learning
- Competent in programming and critical analysis
- Knowledge of security and privacy of machine learning
- Good written and oral communication skills
- Strong motivation, with evidence of independent research skills relevant to the project
- Good time management
Desirable attributes
- Programming experience in Python and Machine Learning frameworks (e.g. TensorFlow or Keras)
- Good knowledge of linear algebra, Bayesian statistics, group theory, etc.
- Experience in Federated Learning and its applications in digital healthcare
Please contact Dr Zhiyuan Tan ([Email Address Removed]) if you are interested in or have any queries about this PhD project.