Machine Learning-Based Antenna Selection in Untrusted Relay Networks

This paper studies the transmit antenna selection based on machine learning (ML) schemes in untrusted relay networks. First, we state the conventional antenna selection scheme. Then, we implement three ML schemes, namely, the support vector machine-based scheme, the naive-Bayes-based scheme, and the k-nearest neighbors-based scheme, which are applied to select the best antenna with the highest secrecy rate. The simulation results are presented in terms of system secrecy rate and secrecy outage probability. From the simulation, we can conclude that the proposed ML-based antenna selection schemes can achieve the same performance without amplification at the relay, or small performance degradation with transmitted power constraint at the relay, comparing with conventional schemes. However, when the training is completed, the proposed schemes can perform the antenna selection with a small computational complexity.