Soumia comes from Algeria and is joining the university of Luxembourg. She holds a double MSC and engineering degrees in computer science. Soumia stands between Machine learning and cyber security and is interested in making data driven solutions privacy friendly.
Project Title: Technologies for algorithms and algorithmic transparency and fairness
Objectives: 1. Investigating the gap between the data protection legislations and the actual technological implementation under complex machine learning pipelines: a. Providing guidelines to implement the transparency tools as stated in the legislations that govern the data workflows. b. Pointing out the failure gaps of the existing legislations Vs the actual nature of the data flows
2. Proposing dynamic methods to set privacy budgets that do not harm the utility of machine learning models and designing a better way of managing and setting privacy budgets.
3. Propose solutions that improve the performance of learning from encrypted data where we preserve the privacy of the data.
4.Designing and discussing frameworks of full privacy preserving machine learning pipeline.
5.Evaluating the privacy risks of implementing transparency and fairness into machine learning pipelines.
6.Proposing and validating performant yet privacy preserving machine learning tools.
- Soumia Zohra El Mestari (2022), “Challenges Of Using Homomorphic Encryption In Machine Learning (Poster)”, in 18th Conference on Artificial Intelligence Applications and Innovations (AIAI2022). 17-20 June, 2022. Crete, Greece.
- Soumia Zohra El Mestari (2022), “Privacy Preserving Machine Learning Systems”, in 5th AAIA / ACM Conference on Artificial Intelligence, Ethics, and Society (AIES2022). August 1-3, 2022. Oxford, United Kingdom.