PhD Researcher from the Information Networks Department Publishes a Scientific Study on a Blockchain-Orchestrated Federated Learning Framework for Intelligent Intrusion Detection
Duhaa Fadill Abbas
As part of the Information Networks Department’s commitment at the College of Information Technology to advancing high-quality scientific research and keeping pace with global developments in artificial intelligence, cybersecurity, and intelligent networking systems, PhD candidate Muthanna Jabbar Abdulridha from the Information Networks Department, under the supervision of Professor Wesam Sameer Bhaya, Dean of the College of Information Technology, has published a specialized scientific research paper presenting an advanced framework for cyber intrusion detection based on federated learning, blockchain technology, and software-defined networking.
The research is entitled:
“A Blockchain-Orchestrated Software-Defined Federated Learning Framework for Intelligent and Decentralized Intrusion Detection”
and was published in the International Journal of Intelligent Engineering and Systems, a reputable journal indexed in Scopus, reflecting the scientific significance of the study and its contribution to addressing contemporary challenges in information security and distributed intelligent systems.
The study addresses several critical limitations associated with conventional federated learning-based intrusion detection systems, particularly the lack of reliable mechanisms for validating the integrity and provenance of client model updates, insufficient integration between auditing and network control layers, and the substantial communication overhead generated by the exchange of model parameters. To overcome these challenges, the researcher proposed an integrated architectural framework, namely BOSD-FD-IDS, which combines blockchain governance mechanisms with Software-Defined Networking (SDN) within a secure and decentralized federated learning environment.
The proposed framework is capable of orchestrating a heterogeneous set of intrusion detection models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory Networks (LSTMs), Gated Recurrent Units (GRUs), and Transformer-based models, through an efficient logit-sharing protocol that eliminates the need to exchange complete model weights. This approach significantly reduces communication costs while maintaining learning efficiency. Furthermore, the framework records training outcomes on a private Ethereum-based blockchain, ensuring data integrity, transparency, and resistance to tampering. Real-time network metrics and trust-aware indicators are utilized to identify suspicious participants, enabling dynamic decisions regarding trust recalibration, participation restriction, or complete isolation from the training process.
This research achievement reflects the Information Networks Department’s strategic vision of promoting advanced interdisciplinary research that integrates artificial intelligence, cybersecurity, and modern networking technologies. It also contributes to the development of innovative solutions capable of addressing emerging digital security challenges while strengthening the College’s role in knowledge production, scientific excellence, and service to both the academic and technological communities.
Research Link:
https://inass.org/wp-content/uploads/2026/03/2026063029-2.pdf