Joint Relay and Jammer Selection Improves the Physical Layer Security in the Face of CSI Feedback Delays

We enhance the physical-layer security (PLS) of amplify-and-forward relaying networks with the aid of joint relay and jammer selection (JRJS), despite the deliterious effect of channel state information (CSI) feedback delays. Furthermore, we conceive a new outage-based characterization approach for the JRJS scheme. The traditional best relay selection (TBRS) is also considered as a benchmark. We first derive closed-form expressions of both the connection outage probability (COP) and of the secrecy outage probability (SOP) for both the TBRS and JRJS schemes. Then, a reliable-and-secure connection probability (RSCP) is defined and analyzed for characterizing the effect of the correlation between the COP and SOP introduced by the corporate source-relay link. The reliability-security ratio (RSR) is introduced for characterizing the relationship between the reliability and security through the asymptotic analysis. Moreover, the concept of effective secrecy throughput is defined as the product of the secrecy rate and of the RSCP for the sake of characterizing the overall efficiency of the system, as determined by the transmit SNR, secrecy codeword rate and the power sharing ratio between the relay and jammer. The impact of the direct source-eavesdropper link and additional performance comparisons with respect to other related selection schemes are further included. Our numerical results show that the JRJS scheme outperforms the TBRS method both in terms of the RSCP as well as in terms of its effective secrecy throughput, but it is more sensitive to the feedback delays. Increasing the transmit SNR will not always improve the overall throughput. Moreover, the RSR results demonstrate that upon reducing the CSI feedback delays, the reliability improves more substantially than the security degrades, implying an overall improvement in terms of the security-reliability tradeoff. Additionally, the secrecy throughput loss due to the second hop feedback delay is more pronounced th- n that of the first hop.