Federated and Privacy-Preserving Learning for Next-Generation Intrusion and Phishing Detection

Machine learning and deep learning have become central to modern intrusion detection and phishing defense, but most deployed systems still rely on centralizing sensitive network and user data to train effective models. This creates real tension between detection performance and data privacy, particularly for organizations bound by regulatory, competitive, or ethical constraints on data sharing. Federated learning offers a path around this trade-off, allowing multiple parties to collaboratively train detection models without exposing raw traffic, logs, or user behavior. Realizing this potential at scale, however, raises open questions around communication efficiency, robustness to non-independent and heterogeneous data across clients, resistance to poisoning and adversarial participants, and how to preserve model interpretability once training is distributed.This Research Topic focuses on advancing federated and other privacy-preserving learning paradigms specifically for intrusion and phishing detection, and on closing the gap between federated learning’s theoretical privacy guarantees and its practical reliability in production security systems. Relevant topics include, but are not limited to:- Federated learning architectures for network intrusion detection across organizational or device boundaries- Privacy-preserving techniques (differential privacy, secure aggregation, homomorphic encryption) applied to phishing and anomaly detection- Robustness of federated models to poisoning, backdoor, and adversarial attacks in security contexts- Handling non-IID and imbalanced threat data across distributed clients- Explainable AI methods compatible with federated or decentralized training- Lightweight and communication-efficient federated models for resource-constrained and IoT environments- Real-time phishing detection using collaborative or federated learning- Benchmarking and evaluation frameworks for federated security models under realistic deployment conditions- Case studies of federated learning deployed in operational cybersecurity or intrusion detection systemsBy bringing together advances in federated learning and applied cybersecurity, this Research Topic aims to establish practical, privacy-preserving foundations for the next generation of collaborative intrusion and phishing detection systems. Topic editors: Pankaj Chandre, Nagesh Jadhav, Amir Javed Source: https://www.frontiersin.org/research-topics/85169/federated-and-privacy-preserving-learning-for-next-generation-intrusion-and-phishing-detection