Neuro-Evolutionary Trust-Driven Federated Learning Framework for Robust Intrusion Detection in Industrial IoT Systems
Journal of Intelligent Decision Making and Information Science · 09.09.2026 · 3 (18s) · 904-926
DOI: 10.17762/jidmis.v3.5764 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 09.09.2026. Verfasst von Ahmed Adnan Hadi. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Industry 4.0, which combines the Internet of Things (IoT) and Artificial Intelligence (AI), supports a wide range of automated operations in modern industries. This study connects multiple industrial IoT (IIoT) devices and uses federated learning over 5G edge networks to support distributed sensing and data analysis. Although Federated Learning (FL) helps protect private data, it remains vulnerable to ransomware attacks, unreliable client contributions, and differences in client data. These issues can affect the reliability of the shared model and make it difficult to maintain consistent intrusion detection across different clients. In addition, varying traffic patterns and limited computing resources at edge devices can reduce the efficiency of the learning process. Therefore, a robust and trust-aware federated learning approach is required to identify malicious or unreliable clients while maintaining privacy-preserving collaboration. To address these issues, this study develops the Neuro-Secure Federated Learning Framework (Neuro_SFL). The CNN–GRU–LSTM model learns spatial and temporal patterns from network traffic for intrusion detection, while FGSM and PGD-based adversarial training improve resistance to manipulated inputs. A temporal trust memory mechanism tracks changes in client trust over time, and the Neuro-Evolutionary Adaptive Gifford Vulture Optimizer (NE-AGVO) improves the aggregation of trust-weighted client updates. Reinforcement learning-based weighting further adjusts the contribution of each client during model training. The proposed Neuro_SFL achieves an accuracy of 99.64%, showing improved intrusion detection performance under different client and attack conditions. Thus, the framework provides a practical approach for improving security and reliability in 5G-enabled IIoT networks.
Zitieren
APA
Hadi, A. A. (2026). Neuro-Evolutionary Trust-Driven Federated Learning Framework for Robust Intrusion Detection in Industrial IoT Systems. Journal of Intelligent Decision Making and Information Science, 3(18s), 904-926. https://doi.org/10.17762/jidmis.v3.5764
Vancouver
Hadi AA. Neuro-Evolutionary Trust-Driven Federated Learning Framework for Robust Intrusion Detection in Industrial IoT Systems. Journal of Intelligent Decision Making and Information Science. 2026;3(18s):904-926. doi:10.17762/jidmis.v3.5764
BibTeX
@article{Hadi2026,
author = {Hadi, Ahmed Adnan},
title = {Neuro-Evolutionary Trust-Driven Federated Learning Framework for Robust Intrusion Detection in Industrial IoT Systems},
journal = {Journal of Intelligent Decision Making and Information Science},
year = {2026},
volume = {3},
number = {18s},
pages = {904--926},
doi = {10.17762/jidmis.v3.5764}
}
EUVIMED