A novel univariate feature selection with ANOVA F-test-based machine learning model for Intrusion Detection Framework of Robotics system

Abstract

Robotic systems have become popular across various industries, ranging from manufacturing and healthcare to logistics and space exploration. However, increasing the integration of robotic systems into critical infrastructures exposes devices to cybersecurity threats. The intrusion detection system (IDS) plays a vital role in safeguarding the systems from malicious activities and unauthorized access. This paper presents a novel, robotics-aware IDS framework incorporating hybrid feature selection and tailored classification strategies for robotic system. To evaluate the efficacy of the presented framework, an algorithm is also designed and tested using multiple machine-learning techniques. The NSL-KDD dataset is utilized for training and evaluating machine learning models due to the inclusion of a wide range of attack scenarios and normal instances. The results demonstrate that the proposed IDS effectively classifies cyberattacks relevant to robotic systems. The presented framework is also evaluated against existing IDS approaches in robotic systems. The results demonstrate that the proposed approach exhibits better results in terms of accuracy, robustness, and adaptability to emerging cyber threats.