5G-Fuzz: An Attack Generator for Fuzzing 5GC, using Generative Adversarial Networks

The evolution of fifth-generation (5G) networks represents a significant technological leap towards a seamless and advanced user experience, allowing hyperconnected use cases with faster data transfer, lower latency and better connectivity for a wide range of mobile devices. In particular, a key element of 5G is the 5G Core (5GC) which follows a service-based architecture, enabling network slicing and improved Quality of Service (QoS). However, despite the benefits of 5GC, it also creates important security and privacy concerns. First, 5GC can combine heterogeneous technologies that can increase the growing attack surface. On the other hand, 5G handles a vast amount of sensitive data that may reflect an attractive goal for potential cyberattackers. Based on the previous remarks, in this paper, we introduce 5G-Fuzz. 5G-Fuzz is a smart fuzzer which takes full advantage of historical data in order to fuzz and target the Packet Forwarding Control Protocol (PFCP) communications between the Session Management Function (SMF) and User Plane Function (UPF). For this purpose, two PFCP attacks are used. In contrast to conventional fuzzers, 5G-Fuzz adopts two Generative Adversarial Networks (GANs) in order to identify and generate the appropriate values of Session Endpoint Identifier (SEID) and sequence number (seq) utilised in the PFCP sessions, thus accelerating the PFCP attacks. Finally, 5G-Fuzz composes and replays the malicious PFCP packets against UPF.