ROS2기반 사이버물리시스템 침입탐지시스템
Intrusion Detection System for Cyber-Physical Systems Based on ROS2
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
최근 국방 분야에서 병력 감소 문제와 현대전의 게임 체인저로 드론 및 로봇의중요성이 부각되고 있다. 그러나 이러한 무인 이동체 기술의 확산과 함께 사이버위협에 대한 연구도 지속적으로 이루어져 이에 대한 사이버보안 대책이마련되어야 한다. 이에 본 연구는 무인 이동체의 ROS2(Robot Operating System 2) 기반 네트워크에서 활용 가능한 인공지능 기반 침입 탐지 모델을제안한다. 첫째, DDS(Data Distribution Service)의 데이터 전달 프로토콜인RTPS(Real-Time Publish-Subscribe)의 특성을 활용하여 사이버 공격을탐지하는 방법을 제시한다. 둘째, 오픈 소스 AI 모델인 Llama를 활용해Serialized Data를 해석하지 않고도 사이버 공격을 탐지하는 방법을 제안한다. 모델의 성능 평가는 실제 UGV를 모델화한 ROS2 기반 데이터셋인 ‘RTPS Attack Dataset’에서 수행하였다. LightGBM 모델을 활용한 평가 결과, F1-Score 95%, recall 98%, accuracy 91%의 성능을 보였다.
Abstract
In modern warfare, the importance of drones and robots as game-changers has been highlighted, particularly in addressing the challenge of declining military personnel. However, with the proliferation of unmanned vehicle technologies, the need for cybersecurity measures to mitigate cyber threats is becoming increasingly critical. This paper proposes an AI-based intrusion detection model applicable to ROS2 (Robot Operating System 2) networks of unmanned vehicles. First, it present a method for detecting cyber attacks by leveraging the characteristics of the RTPS (Real-Time Publish-Subscribe) protocol, which serves as the data delivery protocol for DDS (Data Distribution Service). Second, we propose a novel approach that employs the open-source AI model, Llama, to detect cyber attacks without interpreting serialized data. The performance of the proposed model was evaluated using the "RTPS Attack Dataset," a ROS2-based dataset modeled after a real-world UGV (Unmanned Ground Vehicle). Using a LightGBM model for evaluation, the proposed approach achieved an F1-score of 95%, a recall of 98%, and an accuracy of 91%.
참고문헌 (17)
- "국방혁신 4.0"[URL]
- "What is ROS?"[URL]
- "DDS Interoperability Wire Protocol"[URL]
- McClean, J., "A preliminary cyber-physical security assessment of the robot operating system (ros)", Unmanned systems technolgyxv, 8741, pp.341~348 (2013), SPIE
- B. Dieber, "Security for the robot operating system", Robotics and Autonomous Systems, 98, pp.192~203 (2017)
- B. Dieber, "Penetration testing ROS", Robot Operating System(ROS) The Complete Reference Volume, 4, pp.183~225 (2020)
- "Data Distribution Service (DDS)"[URL]
- "About the DDS Interoperability Wire Protocol Specification Version 2.5"[URL]
- "Introducing Llama 3.2"[URL]
- "meta-llama/Llama-3.2-1B"[URL]
- SONG, Hyun Min, "Intrusion detection system based on the analysis of time intervals of CAN messages for in-vehicle network", 2016international conference on information networking (ICOIN), pp.63~68 (2016), IEEE
- Finder, Ido, "Time-interval temporal patterns can beat and explain the malware", Knowledge-Based Systems, 241, pp.108266~ (2022)
- Hossain, Md Delwar, "LSTM-based intrusion detection system for in-vehicle can bus communications", Ieee Access, 8, pp.185489~185502 (2020)
- JEONG, Seonghoon, "MUVIDS: false MAVLink injection attack detection in communication for unmanned vehicles", Workshop on automotive and autonomous vehicle security (AutoSec), pp.25~ (2021)
- DEĞIRMENCI, Elif, "ROSIDS23: Network intrusion detection dataset for robot operating system", Data in Brief, pp.109739~ (2023)
- ZAFAR, Muhammad Hamza, "Enhanced Intrusion Detection in Robot Operatin g Systems via Grid Search Based Multi-Head Attention Stacked Convolutional Network", 2024 IEEE 20th International Conference on Automation Science and Engineering (CASE), pp.3880~3885 (2024), IEEE