International Journal of Advances in Computer Science & Engineering Research 126 Attack Detection in Cloud Robotics Automation Using Temporal Attention GNNs, VAE-RNN, and Multi- Criteria Decision Analysis for Secure Command Verification Subramanyam Boyapati 1, * , Aravindhan Kurunthachalam 2 1 American Express, Arizona, USA. Email: subramanyam.boyapati86@gmail.com, 2 Associate Professor, School of Computing and Information Technology, REVA University, Bangalore, Karnataka. Email: Aravindhan03@gmail.com Abstract BACKGROUND INFORMATION: Cloud robotics systems are increasingly vulnerable to cyberattacks, which can compromise operational safety and security. Attack detection mechanisms are essential for safeguarding these systems, particularly in cloud-based environments where real-time processing is critical. OBJECTIVES: This paper aims to propose a hybrid attack detection system using Temporal Attention GNNs, VAE-RNN, and Multi-Criteria Decision Analysis (MCDA) for secure command verification in cloud robotics automation, enhancing detection accuracy and scalability. METHODS: The proposed system combines Temporal Attention GNNs for sequential data analysis, VAE-RNN for anomaly detection, and MCDA for decision-making, optimizing attack detection performance while ensuring scalability and efficiency in cloud robotics. RESULTS: Experimental results show that the hybrid model achieves a detection accuracy of 95.2%, with a training time reduction of 30% compared to traditional systems, demonstrating its effectiveness for real-time attack detection in cloud robotics environments. CONCLUSION: The proposed model provides an effective solution for securing cloud robotics automation. It combines cutting-edge deep learning techniques with decision analysis, ensuring robust protection against evolving cyber threats in real-time operations. Keywords: Attack detection, cloud robotics, Temporal Attention GNN, VAE-RNN, MCDA, cybersecurity, anomaly detection, machine learning, command verification, real-time processing. 1. INTRODUCTION The convergence of cloud computing and robotics has sparked a new era of automation to bring about more scalable, flexible, and smarter robotic systems. Cloud robotics utilizes cloud infrastructure to offload computationally expensive operations, save data, and allow robots to use shared resources in real-time. This idea has revolutionized sectors like manufacturing, healthcare, and logistics, augmenting the intelligence and capabilities of *Corresponding Author Name: Subramanyam Boyapati Email: subramanyam.boyapati86@gmail.com robots with the vast computational capacity and data handling abilities of the cloud (Narla, 2024 [1]; Devarajan, 2023 [2]; Valivarthi, 2024 [3]; Peddi et al., 2018 [4]. Nevertheless, such as with any connected system, cloud robotics faces significant cybersecurity challenges. The complexities of such challenges are compounded by the complex interactions between physical robotic systems, cloud platforms, and communication networks (Devarajan et al., 2024 [10]; Valivarthi, 2023 [12]; Narla, 2022[9]). A main issue is the detection of cyber-attacks for cloud-based robotic systems Peddi et al., (2019) [14]. Cyber-attacks can cause severe malfunctioning in robotic automation, modify commands, or halt data transfer between robots and the cloud with the possibility of causing catastrophic consequences in industrial or medical applications