An Ensemble Approach for Fake Base Station Detection using Temporal Graph Analysis and Anomaly Detection
Sheng Sun, Ibrahum Abualhaol, Gwenael Poitau, Ali A. Esswie, Morris Repeta · 2024
The Fake Base Station (FBS) attack presents one of the top threats to the today’s cellular networks, including the 5G networks, it often pretends to be the legitimate cellular network operations and leads to many types of attacks, such as the identity theft, scamming text messages, fraud web sites etc. Traditional detection methods are usually operated through manual inspection with scanning tools, and analysis of radio signals. They often fail in accurately identifying the fake base stations and the results contain substantial false positives. Some of the sophisticated Fake Base Stations employ the detection evasion techniques, such as intermittently turning off the radio, or moving to another location etc. In this study, we propose an AI/ML approach that combines temporal graph analysis and anomaly detection algorithms to enhance the detection accuracy, and overcome the detection evasion techniques that have been overlooked by other technologies. Our Temporal Graph Isolation Forest & Local Outlier Factors (TGIF & LOF) ensemble algorithm leverages the temporal dynamics of the UE’s connectivity and hand-off events, along with the radio measurement reports regularly submitted to the core networks, to detect the anomalies within networks by implementing the Isolation Forest (IF) and Local Outlier Factor (LOF) anomaly detection algorithms. Experimental results demonstrate the effectiveness of our approach in detecting the FBS with lower false positives than manual inspection methods. This work can help bring the real time fake base station detection with significant operational efficiency and prevent the occurrence of the fake base stations in cellular networks.