How Federated learning can Help Intrusion Detection and Mitigate Privacy Issues in 6G Networks

Authored by: Martel Innovate

As we stand on the brink of the 6G era, the potential for transformative communication technologies seems boundless. 6G promises not only unprecedented speeds and connectivity but also the integration of advanced features like AI-driven network optimization and ultra-reliable low-latency communications. However, these advancements come with significant privacy and security challenges. The proliferation of devices capable of storing personal data, particularly smartphones and IoT devices, within 6G networks will significantly increase. Traditional methods of threat detection hinge on the centralization of data, which amplifies the risk of personal information being compromised, either directly or through statistical analysis.

In this context, Federated Learning (FL) emerges as a promising solution, offering a way to enhance intrusion detection systems while mitigating privacy concerns. By processing data locally and aggregating only essential information, FL promises a significant leap in protecting user data and ensuring network integrity.

Traditional intrusion detection systems are often centralized and while effective to a degree, often struggle against the dynamism and scale of modern cyber threats. HORSE project aims to deploy a distributed, reliable AI Engine utilizing federated learning techniques that enables high-powered computations to occur in a decentralized manner for detecting threats in real-time while preventing the dissemination of sensitive data across the network. The data that resides on client devices, are never shared or transferred but used for local model training. One potential strategy might involve employing centralized federated learning that relies on the usage of a central server that coordinates the learning process, aggregates updates, and distributes the global model. After local training, each client sends its model update to a central server that aggregates these updates using algorithms like Federated Averaging (FedAvg). Then the updated global model is sent back to the client devices, and this cycle repeats until the model performance meets the desired criteria.

The HORSE project also plans to implement Network Data Analytics Function  (NWDAF) as part of its technological advancements. NWDAF is an integral component of modern telecommunication networks, especially within the context of 5G and future generations. NWDAF is designed to collect and analyze data about network performance, user behavior, and service quality, serving as the backbone for intelligent decision-making and resource allocation.

The HORSE project will offer different instances of NWDAF that correspond to different network areas. In this scenario an aggregator NWDAF can serve as a central point for collecting analytics data from other NWDAFs that may cover different service areas and consolidate the analytics information for further processing. Each NWDAF instance is tasked with the collection and consolidation of data across a unique cluster of mobile nodes, as depicted in Figure 1. This setup entails multiple instances for each node cluster, with each instance locally training models using the data at hand. Subsequently, the primary global model, alongside the NWDAF aggregator, takes charge of refreshing the overarching parameters and relaying the updated models to the individual nodes.

Figure 1: Federated Learning architecture designed for HORSE platform

Implementing FL within the context of 6G networks, as undertaken by the HORSE project, is not without its challenges. These include concerns about statistical heterogeneity, communication efficiency, ensuring the scalability of the FL framework to accommodate the vast number of devices in 6G networks, addressing computational constraints, and safeguarding against potential security vulnerabilities within the FL process itself. The HORSE project is committed to tackling these issues head-on, developing innovative solutions that pave the way for the safe and effective deployment of FL in 6G networks.


Liu, Yi, et al. “A secure federated learning framework for 5G networks.” IEEE Wireless Communications 27.4 (2020): 24-31.

Sarhan, Mohanad, et al. “Cyber threat intelligence sharing scheme based on federated learning for network intrusion detection.” Journal of Network and Systems Management 31.1 (2023): 3.

Sirohi, Deepika, et al. “Federated learning for 6G-enabled secure communication systems: a comprehensive survey.” Artificial Intelligence Review (2023): 1-93.

Saeed, Mamoon M., et al. “Anomaly detection in 6G networks using machine learning methods.” Electronics 12.15 (2023): 3300.