Feature-based multi-user authentication for parallel uplink transmissions

We study a multi-user up-link scenario where an attacker tries to impersonate the legitimate transmitters. We present a new framework for deriving a posteriori attack proba- bilities from the channel observations at the access point, which enables fast intrusion detection and authentication at the physical layer and can be exploited to reduce the security overhead by offloading higher-layer authentication schemes. This is highly relevant for delay-sensitive applications that are targeted in 5G where the security overhead may limit the real-time performance. We take a factor-graph approach that can easily be extended to take into account other features, channel models, and radio access schemes. While related works only consider single-link scenarios, the multi-user approach in this paper allows us to exploit the cross-channel correlation of the large-scale fading parameters that is due to the propagation environment for improving the detection performance. As numerical results show, especially for slowly changing channels with high correlation our approach provides significant performance gains.