A trustworthy security model for IIoT attacks on industrial robots

Abstract

The security of industrial Internet of Things (IIOT) has recently attracted significant attention. As typical IIoT systems, industrial robots are suffering from lots of threats involving control, communication, and computing, which are difficult to detect IIoT attacks accurately in real-time due to resource constraints. How to efficiently and accurately identify IoT attacks on industrial robots is challenging. To address this, we propose a trustworthy security model (TSM) with a fusion design that integrates an improved deep Q-network (IDQN) and a control model, thus accelerating the model training process by reducing the network traversal space and improving detection accuracy by establishing a prejudgment mechanism. We initially provide a detailed overview of existing methods for robot security and derive a robot control model consisting of kinematics and kinetic. Then, a 17-labeled dataset named iRobot security dataset is established to train the TSM. Moreover, we established a robot physical platform to evaluate the performance of TSM, and five cyber security indicators are employed to quantify the performance. Experimental results show that TSM detects attacks at an average rate of 98.7 % and a 0.01 s latency, which demonstrates the excellent performance in detection accuracy and detection efficiency, and provides an idea for solving these issues.