<p>The most popular 5G technique for multiple antennas is MIMO (Multiple Input Multiple Output), which improves connectivity, speeds, and user experiences. Non-orthogonal multiple access (NOMA) has been designed as a strategy for improving spectral efficiency while supporting some multiple access interference at receivers in MIMO. Although NOMA is an efficient technology in 5G, it is more susceptible to physical layer attacks compared to present-day technology in cellular systems. One of the main performance bottlenecks in NOMA is the Pilot Contamination Attack (PCA), which prevents users from reaching their full potential in uplink and downlink. The attacker initiates the attack during the channel training phase when he injects the same pilot signal as the authorized users to control the channel estimate result and affect the precoding process. Detecting PCA attacks becomes more challenging in NOMA technology since we can not differentiate between authorized and malicious pilot signals in the same block. In this paper, we present an efficient detection technique for PCA in NOMA by using various statistical measurements for normal traffic as a reference profile, and then Bayes’ theorem is applied to classify any incoming traffic as normal or pilot attack. The simulation results show that the proposed technique succeed to detect the pilot contamination attack with a detection rate of up to 99%.</p>