Private and Efficient Vertical Federated Learning
Beschreibung
Federated Learning (FL) tackles the problem of learning from decentralized data from different entities called clients. In vertical FL, the clients have the same data samples (entries) but different features. However, during the learning, private data can be leaked. In [1] the authors suggest a private VFL framework that uses homomorphic encryption and multi-party computation. Another problem that can arise is that the clients might dropout during the learning. In [2] the authors tackle the privacy and the straggler resiliency problem. This seminar aims to understand VFL settings using privacy-preserving and straggler-resiliency techniques.
References:
[1] - https://arxiv.org/abs/2008.06170
[2] - https://arxiv.org/abs/2304.13407
Kontakt
marvin.xhemrishi@tum.de
Betreuer:
Differentially-Private and Robust Federated Learning
Beschreibung
Federated learning is a machine learning paradigm that aims to learn collaboratively from decentralized private data owned by entities referred to as clients. However, due to its decentralized nature, federated learning is susceptible to poisoning attacks, where malicious clients try to corrupt the learning process by modifying their data or local model updates. Moreover, the updates sent by the clients might leak information about the private data involved in the learning. This thesis aims to investigate and combine existing robust aggregation techniques in FL with differential privacy techniques.
References:
[1] - https://arxiv.org/pdf/2304.09762.pdf
[2] - https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9757841
[3] - https://dl.acm.org/doi/abs/10.1145/3465084.3467919
Voraussetzungen
- Knowledge about machine learning and gradient descent optimization
- Proficiency in Python and PyTorch
- Undergraduate statistics courses
- Prior knowledge about differential privacy is a plus
Kontakt
marvin.xhemrishi@tum.de
luis.massny@tum.de