Abstract: Robotic Operating System (ROS) has emerged as a pivotal middleware for developing applications in modern robotic systems, extending beyond industrial use to various real-world applications. As the adoption of ROS-based systems grows, ensuring their security becomes critical due to the increasing risk of cyber-attacks. Intelligent attack detection frameworks leveraging machine learning have proven effective in mitigating these threats. This research explores advanced attack detection techniques using the ROS cyber-attack dataset and evaluates the performance of multiple machine learning models, including Random Forest, Support Vector Machine (SVM), Naive Bayes, Logistic Regression, K-Nearest Neighbours (KNN), and AdaBoost. Additionally, deep learning architectures, such as Long Short-Term Memory (LSTM) networks and 2D Convolutional Neural Networks (CNN2D), are employed to enhance detection accuracy. Among the evaluated models, CNN2D demonstrates superior performance, leveraging its ability to extract intricate spatial and temporal features from input data. The study highlights the potential of deep learning-based solutions for robust security in ROS-based systems, providing a significant step toward resilient and intelligent attack detection in robotic environments. These findings underscore the importance of integrating advanced detection mechanisms to safeguard the integrity and reliability of robotic systems.

Keywords: ROS Security, Cyber-Attack Detection, Machine Learning, Deep Learning, CNN2D, Robotic Systems.

| DOI: 10.17148/IARJSET.2025.12232