DL-based Attack Classification Framework for Robotic Sensor Communication in Industry 4.0
Khushi Trivedi, Karm Dave, Jay Gor, Rajesh Kumar Gupta, Sudeep Tanwar, Mohsen Guizani · 2025
As industrial robotic systems have become more integral to modern manufacturing, ensuring their cybersecurity and operational efficiency is a crucial part. This paper suggests a multi-layered framework aimed at improving the cybersecurity and operational efficiency of industrial robots. It consists of three layers: the Robot Data Acquisition Layer, which verifies and pre-processes sensor data for safe transmission; the Cybersecurity and Threat Detection Layer, which utilizes AI to classify data as malicious or non-malicious and identify specific types of attacks for targeted countermeasures; and the Data Translation and Execution Layer, which transforms safe data into actionable commands for smooth robot operation. Three models were tested for prediction tasks namely Long Short Term Memory (LSTM), Gated Recurrent Units (GRU) and 1-Dimensional Convolutional Neural Networks (1-D CNN). Using metrics like precision, recall, accuracy, and F1-score the proposed framework performance was measured with an overall accuracy of 97.87% from the experiments. This not only showed an excellent mitigation property against cyberattacks but also robustly improved the data integrity which therefore solved the security threats in industrial automation.