Underlay cognitive networks should follow strict interference thresholds to operate in parallel with primary networks. This constraint limits their transmission power and eventually the area of coverage. Therefore, it is very likely that the underlay networks will make use of relays to transmit signals to the distant secondary users. In this paper, we propose a secondary relay selection scheme which maximizes the end-to-end signal to noise ratio (SNR) for the secondary link while keeping the interference levels to the primary network below a certain threshold. We derive closed form expressions for the probability density function (PDF) of the SNR at the secondary destination, average bit error probability and outage probability. Analytical results are verified through simulations which also give insight about the benefits and tradeoffs of the selective cooperation in underlay cognitive networks. It is shown that, in contrast to non-cognitive selective cooperation, this scheme performs better in low SNR region for cognitive networks.