TY - GEN
T1 - A Lightweight and Responsive On-Line IDS Towards Intelligent Connected Vehicles System
AU - Liu, Jia
AU - Fan, Wenjun
AU - Dai, Yifan
AU - Lim, Eng Gee
AU - Lisitsa, Alexei
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.
PY - 2024/9/9
Y1 - 2024/9/9
N2 - The current intelligent connected vehicles (ICV) system often shares the detected intrusion event to the cloud for further collaborative investigation. The upstream channel leading to the Internet of Vehicles (IoV) cloud is typically vendor-proprietary and costly, and the congestion caused by false alarms even exacerbates the situation. Machine learning (ML) can improve intrusion detection performance by reducing the false alarm rate. However, as a computation-intensive approach, traditional ML is not appropriate for real-time detection. Therefore, this paper proposes a lightweight and responsive on-line intrusion detection approach aiming for the ICV system requiring real-time detection. More specifically, we design a model termed Machine Learning integrated with Blacklist Filter (ML-BF), which leverages the feature engineering and the Bloom filter techniques built on ML to enhance both detection and real-time performances. To evaluate the proposed solution, several experiments are conducted by using the Car-Hacking and CIC-IDS-2017 datasets. The experimental results show that our approach can detect intrusion at a microsecond level with a lower computational cost as well as a lower false positive rate than that in the state-of-the-art.
AB - The current intelligent connected vehicles (ICV) system often shares the detected intrusion event to the cloud for further collaborative investigation. The upstream channel leading to the Internet of Vehicles (IoV) cloud is typically vendor-proprietary and costly, and the congestion caused by false alarms even exacerbates the situation. Machine learning (ML) can improve intrusion detection performance by reducing the false alarm rate. However, as a computation-intensive approach, traditional ML is not appropriate for real-time detection. Therefore, this paper proposes a lightweight and responsive on-line intrusion detection approach aiming for the ICV system requiring real-time detection. More specifically, we design a model termed Machine Learning integrated with Blacklist Filter (ML-BF), which leverages the feature engineering and the Bloom filter techniques built on ML to enhance both detection and real-time performances. To evaluate the proposed solution, several experiments are conducted by using the Car-Hacking and CIC-IDS-2017 datasets. The experimental results show that our approach can detect intrusion at a microsecond level with a lower computational cost as well as a lower false positive rate than that in the state-of-the-art.
KW - Bloom Filter
KW - Intelligent Connected Vehicles
KW - Intrusion Detection
KW - Machine Learning
KW - Responsive Detection
UR - http://www.scopus.com/inward/record.url?scp=85204536552&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-68606-1_12
DO - 10.1007/978-3-031-68606-1_12
M3 - Conference Proceeding
AN - SCOPUS:85204536552
SN - 9783031686054
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 184
EP - 199
BT - Computer Safety, Reliability, and Security - 43rd International Conference, SAFECOMP 2024, Proceedings
A2 - Ceccarelli, Andrea
A2 - Trapp, Mario
A2 - Bondavalli, Andrea
A2 - Bitsch, Friedemann
PB - Springer Science and Business Media Deutschland GmbH
T2 - 43rd International Conference on Safety, Reliability and Security of Computer-based Systems, SAFECOMP 2024
Y2 - 18 September 2024 through 20 September 2024
ER -