Research on Intrusion Detection Method for Industrial Control Systems based on Improved APSO-MKBoost-C Algorithm

• Published in: 2022 IEEE 24th Int Conf on High Performance Computing & Communications; 8th Int Conf on Data Science & Systems; 20th Int Conf on Smart City; 8th Int Conf on Dependability in Sensor, Cloud & Big Data Systems & Application (HPCC/DSS/SmartCity/DependSys) pp. 2221 - 2226
• Main Authors: Li, Xiao, Li, Kewen
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2022
• Subjects: Behavioral sciences; Classification algorithms; Imbalanced data classification; Industrial control; Industrial control systems; Intrusion detection; Machine learning; Machine learning algorithms; Production; Robustness
• DOI: 10.1109/HPCC-DSS-SmartCity-DependSys57074.2022.00328

The networking of industrial control systems improves the industrial production efficiency, but it also suffers from the cyber attack risk. The distribution of intrusion data and normal data is extremely imbalanced, and intrusion data shows strong non-linear features as the amount of intrusion data increases, which increases the difficulty for machine learning algorithms to detect cyber attack behaviors. This paper proposes an intrusion detection method for industrial control systems based on improved APSO-MKBoost-C algorithm. Specifically, the weight adjustment factor is proposed to increase attention to intrusion data. The adaptive inertia weight adjustment strategy is proposed to optimize the weight distribution for kernel classifiers to match the changeable attack behaviors. The standard industrial control system datasets provided by Mississippi State University and UCI repository 2 datasets are used in experiments. Experimental results show that compared with other algorithms, APSO-MKBoost-C algorithm has significant classification performance for imbalanced intrusion data.