3T-ZTFL: A Three-Tier Zero Trust Federated Learning Framework for Privacy-Preserving and Byzantine-Resilient Intrusion Detection in Resource-Constrained IoT Critical Infrastructure

New and rapidly evolving trends in IoT devices that rely on resources have broadened the threat surface beyond the perimeter defenses to include what's known as critical infrastructure. Zero Trust Architecture (ZTA) is one solution to this problem, while federated learning (FL) is the other; both solutions can't be readily integrated with endpoints that have only kilobytes of RAM in storage and intermittent connectivity. We introduce the three-tier Zero Trust federated learning architecture, 3T-ZTFL, which distributes the verification across the device tier, edge tier and coordination tier. The lightweight anomaly detector (368 KB, under our 1MB limit, is based on depth-wise separable convolutions, magnitude pruning and 8-bit quantization), the edge aggregates models updates and implements micro-segmentation policies, and the coordination tier keeps the global model and trust knowledge base. Byzantine resilience is achieved with gradient auditing, and privacy is achieved with a client-side differential privacy which is configurable by the operators. In all four benchmark datasets (N-BaIoT, Edge-IIoTset, CICIoT23 and IoT-23), the percentage of the clients correctly detected is between 97.5% and 98.7%, and can be achieved even with 50% malicious clients. The inference time is on average 82 ms on a Raspberry Pi 4 and the privacy cost (1.1 percentage points of accuracy at ε = 5.0) is negligible. Particularly, gradient auditing can enhance the accuracy of clean data by 1.3–1.4 percentage points, and also be Byzantine robust, without any conflict between the two in terms of performance. The results indicate that such a successful, secure and privacy preserving detection is possible on the hardware representative of the critical infrastructure. There is a need for field validation

Zero Trust Architecture; Federated Learning; IoT Security; Differential Privacy; Byzantine Resilience; Critical Infrastructure Protection

1. Albalawi, O., Alwakeel, A., Alharbi, A., & Aljohani, N. (2026). Lightweight and energy-aware intrusion detection for industrial IoT using TinyML and edge AI. Scientific Reports, 16(1), 18524. [Google Scholar] [Crossref]

2. Almiani, M., AbuGhazleh, A., Al-Rahayfeh, A., Atiewi, S., & Razaque, A. (2020). Deep recurrent neural network for IoT intrusion detection system. Simulation Modelling Practice and Theory, 101, 102031. [Google Scholar] [Crossref]

3. Antonakakis, M., April, T., Bailey, M., Bernhard, M., Bursztein, E., Cochran, J., Durumeric, Z., Halderman, J. A., Invernizzi, L., Kallitsis, M., Kumar, D., Lever, C., Ma, Z., Mason, J., Menscher, D., Seaman, C., Sullivan, N., Thomas, K., & Zhou, Y. (2017). Understanding the Mirai botnet. In Proceedings of the 26th USENIX Security Symposium (pp. 1093–1110). USENIX Association. [Google Scholar] [Crossref]

4. Arimanda, N., Radhakrishnan, R. V., & U., P. (2026). FL-IDS++: A dynamic federated learning framework for intrusion detection with personalized non-IID data, adversarial resilience and energy-efficient lightweight models. Future Generation Computer Systems, 177, 108234. [Google Scholar] [Crossref]

5. Bilal, M. A., & Islam, I. U. (2026). Secure and explainable federated learning for IoT intrusion detection: A comprehensive survey. IEEE Open Journal of the Communications Society, 7, 3650–3679. [Google Scholar] [Crossref]

6. Cheon, J. H., Kim, A., Kim, M., & Song, Y. (2017). Homomorphic encryption for arithmetic of approximate numbers. In T. Takagi & T. Peyrin (Eds.), Advances in Cryptology – ASIACRYPT 2017 (Lecture Notes in Computer Science, Vol. 10624, pp. 409–437). Springer. https://doi.org/10.1007/978-3-319-70694-8_15 Chen, L., Tang, Z., Yang, Y., Xiao, X., & Yang, Y. (2026). Federated deep learning for distributed intrusion detection and privacy preservation in power networks. Scientific Reports. Advance online publication. [Google Scholar] [Crossref]

7. Ferrag, M. A., Friha, O., Hamouda, D., Maglaras, L., & Janicke, H. (2022). Edge-IIoTset: A new comprehensive realistic cyber security dataset of IoT and IIoT applications for centralized and federated learning. IEEE Access, 10, 40281–40306. [Google Scholar] [Crossref]

8. Fernandez, E. B., & Brazhuk, A. (2024). A critical analysis of zero trust architecture (ZTA). Computer Standards & Interfaces, 89, 103832. [Google Scholar] [Crossref]

9. Garcia, S., Parmisano, A., & Erquiaga, M. J. (2020). IoT-23: A labeled dataset with malicious and benign IoT network traffic. Zenodo. https://doi.org/10.5281/zenodo.4743746 Ghimire, S. K., & Lodh, S. (2025). FedSecIoT: A federated learning framework for collaborative and privacy-preserving botnet detection in IoT networks with adversarial robustness. In 2025 IEEE Silchar Subsection Conference (SILCON) (pp. 1–6). IEEE. [Google Scholar] [Crossref]

10. Gumbo, M. L., & Almulhem, A. (2026). Zero trust architecture: A systematic literature review. Journal of Network and Systems Management, 34(1), 25.1–25.72. ✅ Verified (ACM Digital Library, DOI 10.1007/s10922-025-09998-x) [Google Scholar] [Crossref]

11. Gutti, C., Thumula, K., & Balbudhe, P. (2025). Federated learning for distributed IoT security: A privacy-preserving approach to intrusion detection. IEEE Access, 13, 135863–135875. [Google Scholar] [Crossref]

12. Han, S., Mao, H., & Dally, W. J. (2016). Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding. In Proceedings of the International Conference on Learning Representations. [Google Scholar] [Crossref]

13. Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017). MobileNets: Efficient convolutional neural networks for mobile vision applications. arXiv. https://arxiv.org/abs/1704.04861 [Google Scholar] [Crossref]

14. Huma, Z. e., Jan, S. U., Ahmad, J., Buchanan, W., & Pitropakis, N. (2026). Adversarial machine learning in IoT security: A comprehensive survey. ACM Computing Surveys, 58(8), 1–35. [Google Scholar] [Crossref]

15. Jacob, B., Kligys, S., Chen, B., Zhu, M., Tang, M., Howard, A., Adam, H., & Kalenichenko, D. (2018). Quantization and training of neural networks for efficient integer-arithmetic-only inference. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 2704–2713). IEEE. [Google Scholar] [Crossref]

16. Javeed, D., Saeed, M. S., Adil, M., Kumar, P., & Jolfaei, A. (2024). A federated learning-based zero trust intrusion detection system for Internet of Things. Ad Hoc Networks, 162, 103540. ✅ Verified (KB Netherlands, DOI 10.1016/j.adhoc.2024.103540) [Google Scholar] [Crossref]

17. Kumar, L., & Sisodiya, K. S. (2025). Securing industrial IoT with zero trust: A framework for critical infrastructure. In 2025 International Conference on Information, Implementation, and Innovation in Technology (pp. 1–5). IEEE. [Google Scholar] [Crossref]

18. Liu, X., Zhang, Y., & Wang, H. (2024). Zero trust architecture for IoT: A survey. IEEE Internet of Things Journal, 11(5), 7890–7910. [Google Scholar] [Crossref]

19. Marsh, S. P. (1994). Formalising trust as a computational concept [Doctoral dissertation, University of Stirling]. [Google Scholar] [Crossref]

20. McMahan, B., Moore, E., Ramage, D., Hampson, S., & Arcas, B. A. (2017). Communication-efficient learning of deep networks from decentralized data. In Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (pp. 1273–1282). PMLR. [Google Scholar] [Crossref]

21. Meidan, Y., Bohadana, M., Mathov, Y., Mirsky, Y., Shabtai, A., Breitenbacher, D., & Elovici, Y. (2018). N-BaIoT: Network-based detection of IoT botnet attacks using deep autoencoders. IEEE Pervasive Computing, 17(3), 12–22. [Google Scholar] [Crossref]

22. Michalec, O. (2026). Testing and errors: Aligning cybersecurity expertise in contemporary UK energy infrastructures. Science, Technology, & Human Values. Advance online publication. [Google Scholar] [Crossref]

23. Murugan, G., Padmaja, C., Sindhuja, R., Yarde, P., & Sudhakar, M. (2025). A comprehensive survey on cyber-physical system security in the Internet of Things (IoT): Addressing and solutions. Recent Advances in Computer Science and Communications, 18(5). [Google Scholar] [Crossref]

24. Neto, E. C. P., Dadkhah, S., Ferreira, R., Zohourian, A., Lu, R., & Ghorbani, A. (2023). CICIoT2023: A real-time dataset and benchmark for large-scale attacks in IoT environment. Sensors, 23(13), 5941. [Google Scholar] [Crossref]

25. Pathak, R., & Panteli, N. (2026). Governing the agents: Leadership dilemmas in deploying agentic AI in a cyber defence project team. Journal of Information Technology Teaching Cases. Advance online publication. [Google Scholar] [Crossref]

26. Rezaei, H., Taheri, R., Nowroozi, E., Hajizadeh, M., Shiaeles, S., & Bauschert, T. (2025). A survey on security and privacy in federated learning-based intrusion detection systems for 5G and beyond networks. IEEE Open Journal of the Communications Society, 7, 253–300. [Google Scholar] [Crossref]

27. Soomro, I. A., Khan, H. U. R., et al. (2025). SecureDyn-FL: A robust privacy-preserving federated learning framework for intrusion detection in IoT networks. IEEE Transactions on Network and Service Management, 23, 1742–1765. https://doi.org/10.1109/TNSM.2025.3647642 arXiv:2601.06466 [Google Scholar] [Crossref]

28. Spencer, M., & Pizio, D. (2024). The de-perimeterisation of information security: The Jericho Forum, zero trust, and narrativity. Social Studies of Science, 54(5), 655–677. [Google Scholar] [Crossref]

29. Ullah, S., Wu, J., Ahmad, J., Garg, S., Kaddoum, G., & Guizani, M. (2026). Spectre-Fed: Evolving federated edge intelligence from FedEdge-ID to robust-private IoT intrusion detection via hybrid adversarial training. IEEE Open Journal of the Communications Society, 7, 1994–2012. [Google Scholar] [Crossref]

30. Wei, K., Li, J., Ding, M., Ma, C., Yang, H. H., Farokhi, F., Jin, S., Quek, T. Q. S., & Poor, H. V. (2020). Federated learning with differential privacy: Algorithms and performance analysis. IEEE Transactions on Information Forensics and Security, 15, 3454–3469. [Google Scholar] [Crossref]

31. Weyns, M., Ferrero, D., Op de Beek, S., Wagner, D., Smaragdakis, G., & Griffioen, H. (2026). From Mirai to Gorilla: A deep dive into a long-lasting DDoS-for-hire botnet. In Proceedings of the 35th USENIX Security Symposium. [Google Scholar] [Crossref]

• Assessment of the Role of Artificial Intelligence in Repositioning TVET for Economic Development in Nigeria
• Teachers’ Use of Assure Model Instructional Design on Learners’ Problem Solving Efficacy in Secondary Schools in Bungoma County, Kenya
• “E-Booksan Ang Kaalaman”: Development, Validation, and Utilization of Electronic Book in Academic Performance of Grade 9 Students in Social Studies
• Analyzing EFL University Students’ Academic Speaking Skills Through Self-Recorded Video Presentation
• Major Findings of The Study on Total Quality Management in Teachers’ Education Institutions (TEIs) In Assam – An Evaluative Study