PUKYONG

A Novel Intrusion Detection System in IoT Networks Leveraging Blockchain-Enabled Federated Learning

Metadata Downloads
Abstract
The significance of the Internet of Things (IoT) in the newly-revolutionized world cannot be overlooked where we are observing improved IoT features and applications every day. There are currently increased concerns about cyberattacks as IoT networks employ more users and applications. In recent years, the use of Intrusion Detection Systems (IDS) based on machine learning has grown to tackle those cyberattacks. But, due of the high cost of computation and privacy issues, adopting centralized machine learning methods (ML) is not a practical option because there is so much data being saved at one central cloud server. The most promising approach in tackling this challenging problem is federated learning (FL), as it distributes learning to the end devices without sharing private data to the central server. Blockchain (BC) can offer additional benefits if it’s applied in combination with federated learning due to their improved level of privacy and security. In this thesis, a blockchain-based architecture is suggested to enable federated learning in enhancing the intrusion detection of IoT systems. All interactions and transactions in our suggested solution are governed and tracked by Ethereum Smart Contracts (SC), and we also developed a decentralized storage system integrating InterPlanetary File System (IPFS) and Hyperledger Fabric to enhance the security of model learning and findings. The experimental outcomes demonstrate that the suggested federated learning-based method is competitive with the traditional centralized models in its ability to detect unwanted intrusions. The security and privacy of the entire IoT network are also guaranteed by the blockchain's features.|새롭게 혁신된 세상에서 사물 인터넷(IoT)은 매일 개선된 기능과 애플리케이션을 관찰하고 있는 곳에서 그 중요성을 간과할 수 없다. 현재 IoT 네트워크가 더 많은 사용자와 애플리케이션을 사용함에 따라 사이버 공격에 대한 우려가 증가하고 있다. 최근 몇 년 동안 기계 학습(machine learning)을 기반으로 하는 침입 탐지 시스템(Intrusion Detection Systems: IDS)의 사용은 이러한 사이버 공격에 대처하기 위해 성장하였다. 그러나 높은 계산 비용과 개인 정보 보호 문제로 인하여 중앙 집중식 기계 학습 방법은 하나의 중앙 클라우드 서버에 저장되는 데이터가 너무 많기 때문에 실용적인 선택이 아니다. 이 어려운 문제를 해결하는 가장 유망한 접근 방식은 중앙 서버에 개인 데이터를 공유하지 않고, 최종 장치에 학습을 배포하는 연합 학습(federated learning: FL) 기법이다. 블록체인은 향상된 개인 정보 보호 및 보안 수준으로 인하여 연합 학습과 함께 적용할 경우 보다 많은 이점이 제공될 수 있다. 본 논문에서는 IoT 시스템의 침입 탐지를 향상시키기 위하여 연합 학습이 가능한 새로운 블록체인 기반 아키텍처를 제안한다. 제안한 솔루션의 모든 상호 작용 및 트랜잭션은 이더리움 스마트 계약(Ethereum Smart Contracts: SC)에 의해 관리되고 추적되며, 모델 학습 및 보안을 강화하기 위해 IPFS(InterPlanetary File System)와 Hyperledger Fabric을 통합하는 분산 스토리지 시스템도 개발한다. 실험 결과, 제안한 연합 학습 기반 방법은 원하지 않는 침입을 감지하는 능력에서 기존의 중앙 집중식 모델과 경쟁할 수 있음을 보여주었다. 블록체인의 기능은 전체 IoT 네트워크의 보안 및 프라이버시도 보장된다.
Author(s)
Rashid Md Mamunur
Issued Date
2023
Awarded Date
2023-02
Type
Dissertation
Keyword
Internet of Things, Intrusion Detection System, Federated Learning, Blockchain, Security, Privacy
Publisher
부경대학교
URI
https://repository.pknu.ac.kr:8443/handle/2021.oak/32882
http://pknu.dcollection.net/common/orgView/200000663739
Affiliation
Pukyong National University, Graduate School
Department
대학원 인공지능융합학과
Advisor
Ki-Ryong Kwon
Table Of Contents
I. Introduction 1
1.1Overview 1
1.2 Intrusion Detection Systems (IDS) and Machine Learning 2
1.3 Intrusion Detection Systems (IDS) and Federated Learning 4
1.4 Intrusion Detection Systems (IDS) with Blockchain and Federated Learning 5
1.5 Thesis Objectives 7
1.6 Thesis Organization 8
II. Related Works 9
2.1 Overview 9
2.2 Intrusion Detection in IoT using ML 9
2.3 Intrusion Detection in IoT using FL 10
2.4 Intrusion Detection using Blockchain 12
III. Proposed Method 15
3.1 Overview 15
3.2 System Architecture 15
3.2.1 IoT Client and Device Layer for Local Training 16
3.2.2 Smart Contract Layer for FL-Distribution 17
3.2.3 Blockchain Layer for FL Aggregation and Storage 17
3.3 Adversary Model and Assumptions 18
3.4 FL Classifiers for intrusion detection 19
3.5 ML Classifiers for intrusion detection 22
3.5.1 Convolutional Neural Network (CNN) 22
3.5.2 Recurrent Neural Network (RNN) 23
IV. Experimental Setup, Results, and Discussion 26
4.1 Overview 26
4.2 Experimental Setup 26
4.3 Dataset Description and Pre-Processing 26
4.4 Evaluation Metrices 28
4.5 Performance Evaluation 29
4.5.1 Intrusion Detection using Centralized Method 29
4.5.2 Intrusion Detection using Blockchain-enabled FL Method 30
4.6 Deployment of Smart Contract for FL-Distributor 31
4.6.1 FL-clients Registration Smart Contract 31
4.6.2 Local Learning Upload Smart Contract 32
4.6.3 Global Learning Distribution Smart Contract 33
4.6.4 Cost Analysis of Smart Contracts 33
4.6.5 Security Analysis of Smart Contracts 34
4.7 Comparison with Similar Works 35
4.8 Discussion 35
4.8.1 Blockchain-enabled Security Features 35
4.8.2 Benefits of Using Federated Learning in Intrusion Detection 36
V. Conclusion and Future Scope 38
References 39
Acknowledgement 46
Degree
Master
Appears in Collections:
대학원 > 인공지능융합학과
Authorize & License
  • Authorize공개
  • Embargo2023-02-08
Files in This Item:

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.