A robust federated learning and deep learning based cyberattack detection algorithm

Mehmet Akif Alper

doi:10.1038/s41598-026-66336-0

A robust federated learning and deep learning based cyberattack detection algorithm

2026-09-07

0

OA

AI

M

Mehmet Akif Alper *

DOI：10.1038/s41598-026-66336-0

Original

Original request for help

Save

Abstract

En 中文

Distributed Denial-of-Service (DDoS) attacks represent a rapidly growing threat to modern network infrastructures, prompting the development of numerous detection techniques aimed at mitigating their adverse effects. Although Machine Learning (ML) and Deep Learning (DL) approaches have demonstrated strong effectiveness in identifying such attacks, they often suffer from high computational complexity and suboptimal efficiency–accuracy trade-offs. Traditional centralized ML based intrusion detection systems require network aggregation for traffic data, data privacy, and scalability problems. In this context, this research proposes a novel and articulated DL based framework that improves detection performance while significantly reducing computational overhead on local devices for DDoS attack detection. The performance of the proposed algorithm tested using the CIC-DDoS2019 dataset. Based on results on the public dataset, our proposed algorithm presents improved detection architecture.

A robust federated learning and deep learning based cyberattack detection algorithm

Abstract

Journal

Organization