Realistic Secrecy Performance Analysis for LiFi Systems

This paper studies the secrecy performance of light-fidelity (LiFi) networks under the consideration of random device orientation and partial knowledge of the eavesdroppers' channel state information. Particularly, the secrecy capacity and secrecy outage probability are analysed for the case of a single eavesdropper as well as for the case of multiple eavesdroppers. Moreover, a machine learning based access point (AP) selection algorithm is presented with the objective of maximising the secrecy capacity of legitimate users. Our results show that optimising the AP selection while taking into account the random behaviour of the optical channel results in a significant enhancement in the achievable secrecy performance. In fact, using the derived realistic secrecy expressions as the basis for AP selection results in up to 30% secrecy capacity enhancement compared to the limited assumption of fixed orientation.