사물인터넷 공격 탐지에 대한 연방 학습 방법의 성능 분석

A Robust Federated Learning Approach for Combating Attacks Against IoT Systems Under non-IID Challenges Eyad Gad∗1, Zubair Md Fadlullah∗2, and Mostafa M. Fouda‡§3 ∗Department of Computer Science, University of Western Ontario, London, ON, Canada. ‡Department of Electrical and Computer Engineering, Idaho State University, Pocatello, ID, USA. §Center for Advanced Energy Studies (CAES), Idaho Falls, ID, USA. Emails: 1egad@uwo.ca, 2zfadlullah@ieee.org, 3mfouda@ieee.org Abstract—In the context of the growing proliferation of user devices and the concurrent surge in data volumes, the complex- ities arising from the substantial increase in data have posed formidable challenges to conventional machine learning model training. Particularly, this is evident within resource-constrained and security-sensitive environments such as those encountered in networks associated with the Internet of Things (IoT). Federated Learning has emerged as a promising remedy to these challenges by decentralizing model training to edge devices or parties, effectively addressing privacy concerns and resource limitations. Nevertheless, the presence of statistical heterogeneity in non- Independently and Identically Distributed (non-IID) data across different parties poses a significant hurdle to the effectiveness of FL. Many FL approaches have been proposed to enhance learning effectiveness under statistical heterogeneity. However, prior studies have uncovered a gap in the existing research land- scape, particularly in the absence of a comprehensive comparison between federated methods addressing statistical heterogeneity in detecting IoT attacks. In this research endeavor, we delve into the exploration of FL algorithms, specifically FedAvg, FedProx, and Scaffold, under different data distributions. Our focus is on achieving a comprehensive understanding of and addressing the challenges posed by statistical heterogeneity. In this study, We classify large-scale IoT attacks by utilizing the CICIoT2023 dataset. Through meticulous analysis and experimentation, our objective is to illuminate the performance nuances of these FL methods, providing valuable insights for researchers and practitioners in the domain. Index Terms—Internet of Things (IoT), Federated Learning (FL), non-IID data, statistical heterogeneity, Intrusion Detection System (IDS), scaffolding technique. I. INTRODUCTION In the landscape of Internet of Things (IoT) networks, generating extensive datasets has become a crucial component of contemporary technological progress. The significance of data as a novel production medium is underscored by the vast possibilities enabled through data sharing and mining [1]–[4]. Notably, the potential benefits are evident in scenarios such as the sharing of vehicle IoT data to alleviate congestion and the collaborative analysis of medical big data to facilitate disease prediction [5], [6]. However, the conventional methods of data sharing, which involve direct uploads to centralized reposito- ries, prove insufficient in meeting the escalating demands for robust data security and privacy protection. This inadequacy becomes particularly pronounced when considering the sus- ceptibility of the network traffic and transmitted data across IoT devices to various cyber-attacks. In securing IoT networks against evolving these attacks, traditional machine learning (ML) has been utilized in Intrusion Detection Systems (IDS) to maintain network integrity. However, the computational limitations of IoT devices pose challenges for implementing effective ML-based IDS. The constrained computing capabil- ities of IoT devices make it difficult to process the substantial data volumes required for IDS model training. Additionally, privacy concerns arise with centralized learning approaches in IoT networks, as they involve collecting data or network traffic, potentially compromising sensitive information [7]– [9]. Balancing privacy and effective intrusion detection are essential for establishing trust in IoT systems. To address these challenges, Federated learning (FL) is proposed by decentralizing model training to edge devices or parties, pro- viding an innovative solution that enables collaborative model training without exposing raw data to a central server. FL has attracted many research interests [10]–[18] and been widely used in practice [19]–[22]. This enhances privacy and security and leverages collective network intelligence for improved intrusion detection accuracy and efficiency [23]–[27]. One common challenge in FL arises from statistical het- erogeneity [28]–[31], which encompasses variations in data distribution and characteristics among local datasets, includ- ing differences in class balance or data quality, commonly referred to as non Independently and Identically Distributed (non-IID). Yet, addressing this heterogeneity proves especially challenging in securing IoT networks. The varied distributions of IoT attacks [32] underscore the significance of

…(Full text truncated)…

This content is AI-processed based on ArXiv data.