A method for improving the robustness of federal learning systems based on input transformation

Ziting Gao; Zesan Liu; Chenghua Fu

doi:10.1117/12.2661042

16 December 2022 A method for improving the robustness of federal learning systems based on input transformation

Ziting Gao, Zesan Liu, Chenghua Fu

Proceedings Volume 12500, Fifth International Conference on Mechatronics and Computer Technology Engineering (MCTE 2022); 125005Z (2022) https://doi.org/10.1117/12.2661042
Event: 5th International Conference on Mechatronics and Computer Technology Engineering (MCTE 2022), 2022, Chongqing, China

Abstract

Federated learning is an effective method to solve the problem of data silos, but adversarial attacks launched based on adversarial samples pose a great threat to the security of federated learning models. This makes the application and promotion of federated learning somewhat affected. Therefore, this paper verifies the performance of a defense method for adversarial attacks in federated learning scenario, which is proposed in the traditional machine learning. The method defends adversarial attacks mainly by performing an input transformation before feeding images to the model. We conducted experiments on the EMnist dataset, and the experimental results show that this defense strategy can also improve the robustness of federation learning under different adversarial attacks.

Citation Download Citation

Ziting Gao, Zesan Liu, and Chenghua Fu "A method for improving the robustness of federal learning systems based on input transformation", Proc. SPIE 12500, Fifth International Conference on Mechatronics and Computer Technology Engineering (MCTE 2022), 125005Z (16 December 2022); https://doi.org/10.1117/12.2661042

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available