SHIFT: Enhancing federated learning robustness through client-side backdoor detection

doi:10.1016/j.inffus.2026.104144

SHIFT: Enhancing federated learning robustness through client-side backdoor detection

2026-01-10

0

PRE

AI

K

Kang Wang

王亮亮 (Liangliang Wang)

DOI：10.1016/j.inffus.2026.104144

Original

Original request for help

Save

Abstract

En 中文

• SHIFT reduces server-side computational load by shifting backdoor detection to the client side. • SHIFT significantly improves the efficiency of backdoor detection by 28 to 36.65 times compared to existing schemes. • SHIFT enhances security by employing client-side code obfuscation and dynamic risk mapping to prevent malicious clients from bypassing detection.

SHIFT: Enhancing federated learning robustness through client-side backdoor detection

Abstract

Journal

Organization