Enhancing Intrusion Detection Systems for 5G Networks using AI
Abirami Gurushanker, Kirtanaa Anandakumaran, C. Christopher Columbus · 2024
As 5G networks advance, cybersecurity vulnerabilities in IoT devices are increasingly exposed. While intrusion detection systems (IDS) have progressed, there is a notable gap in IDS specifically designed for the unique protocols and complexities of 5G networks. Current research often emphasises binary classification, missing the nuanced multiclass classification needed for effective threat detection. This study addresses these deficiencies by integrating machine learning and deep learning techniques within the 5G core architecture to develop an advanced IDS. Utilising TCP/IP flow statistics and PFCP signalling attack data, and focusing on control plane signalling between the Session Management Function (SMF) and User Plane Function (UPF), various models are evaluated for detecting anomalous activities. The study achieves 97% accuracy with TCP/IP data and 69% with PFCP data. The proposed IDS, tailored for 5G-specific workloads and trained on PFCP protocol data, enhances detection capabilities and addresses significant security challenges, improving the resilience of IoT devices in 5G networks.