Attacks on federated learning and defense strategy
Smartphone devices have become part of our daily lives, and with a simple touch, users are contributing to Machine Learning (ML). Federated Learning is new privacy-preserving collaborative learning technique that addresses the problem of conventional ML. Nevertheless, it still has a broad attack sur...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2021
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/153254 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Smartphone devices have become part of our daily lives, and with a simple touch, users are contributing to Machine Learning (ML). Federated Learning is new privacy-preserving collaborative learning technique that addresses the problem of conventional ML. Nevertheless, it still has a broad attack surface area and is vulnerable to adversarial attacks. The project aims to create a system to aid in Federated attack discovery and research.
The report covers the design, implementation, testing and evaluation of the proposed system. Federated Learning Attack Simulation (FLAS) system will be the first website application design with an easily operated workflow for non-experts and incorporates the capabilities to speed up testing and analysis for Federated Learning (FL) professionals. The project has also been submitted to the AAAI-22 conference and is undergoing review.
Although FLAS will be beneficial to research, it is still in the early stage of development. Future studies could explore how to incorporate simulations of other federated attacks and allow more diverse federated model training and analysis. |
|---|