| Details |
|---|
| Project Title |
Securing Federated Learning Through Fully Homomorphic Encryption
|
| Grant |
Korean National Research Fund |
|
|
| Campus |
Kampar campus |
| Stipend |
RM2,500 per month
|
| Entry Requirement |
A candidate with good knowledge of C, C++ or Go language, or CUDA programming will have an advantage. However, this is not a hard requirement.
Note that training in CUDA programming will be provided.
|
| Job Description |
Federated learning (FL) is an emerging research topic that has attracted a lot of attention recently. FL allows multiple clients to collaborate and learn a global model without sharing their raw data. On one hand, this allows a privacy protection on the data, but on the other hand, it leaks information due to the transmission of sensitive data (i.e., model parameters) over insecure communication channels. These concerns can be addressed using fully homomorphic encryption (FHE). However, FHE is known to have slow execution speed, thus requiring performance optimization before it can be used in practical applications.
In this project, we aim to propose a FHE protected FL system with high performance. This can be achieved through GPUs as accelerators, and some novel techniques to pack/unpack data to allow efficient parallel implementation.
The candidate is required to perform one of the followings:
- Implement a prototype FL system for a specific application
- Interface the FL system with existing FHE libraries (e.g., OpenFHE, PhantomFHE)
- Utilize GPU acceleration to speed up the FL system
- Propose some innovative applications using FL or deep learning
|
| Field of study |
Computer Science (Master of Science (Computer Science))
|
| How to Apply |
If you are interested, please contact Dr Lee Wai Kong via email at wklee@utar.edu.my
|
