FedSec: Advanced Threat Detection System for Federated Learning Frameworks
K.P.S Avishka, G.K.A Gunasekara, K.K Maduwantha, J.P.L.C Srilal, Kanishka Yapa, Laneesha Ruggahakotuwa · 2023
Federated learning (FL) is a machine learning approach that enables the training of models across multiple decentralized devices or servers while keeping the training data local. While FL offers advantages such as privacy preservation and decentralized data training, it also introduces certain security and privacy challenges. The detection of attacks in the context of FL holds significant importance as it serves as a critical safeguard for the integrity, security, and effectiveness of the collaborative training process. This paper primarily centers its attention on the detection of specific types of attacks, specifically addressing model poisoning, data poisoning, and the Sybil attack. Distinct methodologies are employed for the detection of each specific attack. To achieve this objective, technologies such as min-max scaling, Generalized ESD testing, and the Cauchy mean value theorem are utilized. Upon successful detection, it triggers an alert and presents a notification. The related outcome is exhibited in a dedicated dashboard called ‘FedSec’.