Summary This paper evaluates the secrecy performance for the amplify‐and‐forward (AF) protocol with partial relay selection (PRS) schemes in the scenario of multiple independent but not necessarily identically distributed eavesdroppers. The secrecy performances of the first‐hop and second‐hop PRS schemes are revisited. Given the secrecy data rate, an efficient PRS scheme is presented, which selects the relay based on the instantaneous channel state information (CSI) of either the first or second hop according to the statistical CSI of two hops. The proposed PRS scheme provides the trade‐off between the two‐hop and one‐hop PRS criteria. Results show that the secrecy outage probability (SOP) performance of the proposed scheme is close to that of the conventional opportunistic relaying in the medium main‐to‐eavesdropper ratio (MER) regions. As for the intercept probability, the second‐hop scheme achieves the highest diversity order among PRS schemes and its performance depends on the overall eavesdroppers' behavior not on individual effects of the specific relay. The proposed criterion also outperforms the first‐hop PRS scheme. Simulation results finally substantiate the accuracy of the theoretical analysis.