LSTM-Based Early Anomaly Detection for Industrial Robotic Arm Cybersecurity
DOI:
https://doi.org/10.58445/x1e5yf88Keywords:
LSTM autoencoder, anomaly detection, robotic arm, industrial cybersecurity, RoAD DatasetAbstract
Robotic systems have become an increasingly attractive target for attackers, which leads to major losses and disruptions. Manufacturing has become the most targeted industry, with around 27.7% of observed incidents in 2025 (IBM, 2026), and the global average cost of a data breach was $4.44 million (IBM, 2025). Companies continue to face vulnerabilities in their systems, causing shutdowns and damaging their reputation, despite investing billions in robotics. These risks are extremely critical as robotic systems operate as cyber-physical systems that integrate software, sensors, and communication networks. This means that cyberattacks are able to directly impact physical operations and safety (Fosch-Villaronga & Mahler, 2021). The goal of this project is to explore how an LSTM neural network can improve early anomaly detection and reduce cybersecurity threats in robotic systems to minimize both direct and indirect losses. To do this, I developed and tested an AI-driven anomaly detection framework that can monitor robotic behaviour and look out for unusual patterns in behaviour, which is crucial for industrial environments where abnormal sensor behaviour or system failures can cause damage in manufacturing processes and lead to significant operational damage (Lu et al., 2022). This project will use the RoAD dataset to evaluate an LSTM-based autoencoder for detecting anomalies in robotic arm behaviour. The model achieved an AUROC of 0.71 on the collision subset, which is slightly above the best unsupervised result reported with the dataset (Mascolini et al., 2023), though low precision shows that reconstruction error carries only a limited anomaly signal. The weight and velocity subsets are labeled anomalous throughout, so they cannot be used to judge detection ability.
References
Alexandrov, A. (2024). LSTM-RNN method for anomaly-based intrusion detection systems. BISEC’2024: 15th International Conference on Business Information Security. CEUR Workshop Proceedings. https://ceur-ws.org/Vol-3971/paper02.pdf
Alruwaili, A., Islam, S. M. N., & Gondal, I. (2025). Cybersecurity in robotic autonomous vehicles: Machine learning applications to detect cyber attacks. CRC Press. https://www.routledge.com/Cybersecurity-in-Robotic-Autonomous-Vehicles-Machine-Learning-Applications-to-Detect-Cyber-Attacks/Alruwaili-Islam-Gondal/p/book/9781003610915
Canizo, M., Triguero, I., Conde, A., & Onieva, E. (2019). Multi-head CNN–RNN for multi-time series anomaly detection: An industrial case study. Neurocomputing, 363, 246–260. https://doi.org/10.1016/j.neucom.2019.07.034
Fosch-Villaronga, E., & Mahler, T. (2021). Cybersecurity, safety and robots: Strengthening the link between cybersecurity and safety in the context of care robots. Computer Law & Security Review, 41, 105528. https://doi.org/10.1016/j.clsr.2021.105528
IBM. (2025). Cost of a data breach report 2025. IBM Security. https://www.ibm.com/reports/data-breach
IBM. (2026). X-Force threat intelligence index 2026. IBM. https://www.ibm.com/reports/threat-intelligence
Kayan, H. (2025). Anomaly detection in industrial robotic arms using edge-based IoT systems [Doctoral dissertation, Cardiff University]. https://orca.cardiff.ac.uk/id/eprint/179739/
Lu, H., Du, M., Qian, K., He, X., & Wang, K. (2022). GAN-based data augmentation strategy for sensor anomaly detection in industrial robots. IEEE Sensors Journal, 22(18), 17464–17474. https://doi.org/10.1109/JSEN.2021.3069452
Mascolini, A., Gaiardelli, S., Ponzio, F., Dall’Ora, N., Macii, E., Vinco, S., Di Cataldo, S., & Fummi, F. (2023). Robotic Arm Dataset (RoAD): A dataset to support the design and test of machine learning-driven anomaly detection in a production line. IECON 2023 – 49th Annual Conference of the IEEE Industrial Electronics Society. IEEE. https://doi.org/10.1109/IECON51785.2023.10311726
Park, Y., & Yun, I. D. (2018). Fast adaptive RNN encoder–decoder for anomaly detection in SMD assembly machine. Sensors, 18(10), 3573. https://doi.org/10.3390/s18103573
Zhou, K., Cao, Y., Kim, T., Zhao, H., Dong, H., Ting, K. M., & Zhu, Y. (2024). RAD: A dataset and benchmark for real-life anomaly detection with robotic observations. arXiv. https://arxiv.org/abs/2410.00713
Zhou, X., Hu, Y., Liang, W., Ma, J., & Jin, Q. (2021). Variational LSTM-enhanced anomaly detection for industrial big data. IEEE Transactions on Industrial Informatics, 17(5), 3469–3477. https://doi.org/10.1109/TII.2020.3022432
Downloads
Posted
Categories
License
Copyright (c) 2026 Research Archive of Rising Scholars
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
