Research on a PSO-H-SVM-Based Intrusion Detection Method for Industrial Robotic Arms

• Published in: Applied sciences Vol. 12; no. 6; p. 2765
• Main Authors: Zhou, Yulin, Xie, Lun, Pan, Hang
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.03.2022
• Subjects: Algorithms; Artificial intelligence; Communication; Design; Expected values; Fault diagnosis; Global positioning systems; GPS; intrusion detection; Logic; Machine learning; Manufacturing; physical process logic attack; PSO-H-SVM; robotic arm system; Robotics; Robots; security; Security systems; Sensors; Support vector machines
• DOI: 10.3390/app12062765
• ISSN: 2076-3417

The automation and intelligence of industrial manufacturing is the core of the fourth industrial revolution, and robotic arms and proprietary networked information systems are an integral part of this vision. However, with the benefits come risks that have been overlooked, and robotic arms have become a heavily attacked area. In order to improve the security of the robotic arm system, this paper proposes an intrusion detection method based on a state classification model. The closure operation process of the robotic arm is divided into five consecutive states, while a support vector machine based on the particle swarm optimization algorithm (PSO-H-SVM) classifies the operation state of the robotic arm. In the detection process, the classifier predicts the operation state of the robotic arm in real time, and the detection method determines whether the state transfer meets the logical requirements, and then determines whether the intrusion occurs. In addition, a response mechanism is proposed on the basis of the intrusion detection system to make protection measures for the robotic arm system. Finally, a physical experiment platform was built to test the intrusion detection method. The results showed that the classification accuracy of the PSO-H-SVM algorithm reached 96.02%, and the detection accuracy of the intrusion detection method reached 90%, which verified the effectiveness and reliability of the intrusion detection method.