Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about FBSDetector, a machine learning-based security solution designed to detect fake base stations and multi-step attacks in cellular networks through this 11-minute conference presentation from USENIX Security '25. Discover how researchers from Purdue University developed an effective detection system that operates at the user equipment side using layer-3 network traces, achieving 96% accuracy in fake base station detection with only a 2.96% false positive rate. Explore the creation of FBSAD and MSAD, the first comprehensive datasets incorporating instances of fake base stations and 21 multi-step attacks captured across various real-world cellular network scenarios including mobility and different attacker capabilities. Understand the novel multi-level machine learning framework that employs stateful LSTM with attention mechanisms for packet classification and graph learning for multi-step attack recognition, demonstrating 86% accuracy in attack detection with a 3.28% false positive rate. Examine how this solution addresses critical security threats including unauthorized surveillance, sensitive information interception, and network service disruption caused by malicious base stations impersonating legitimate ones. See how FBSDetector has been deployed as a real-world mobile application solution and validated in live environments, offering significant improvements over existing heuristic-based approaches that fail to detect fake base stations in practical scenarios.
