In this work, we investigate the physical layer security of a jamming-based underlay cognitive hybrid satellite-terrestrial network {consisting of a} radio frequency link at the first hop and an optical feeder at the second hop. Particularly, one secondary user (SU) is transmitting data {to an end-user} optical ground station {($\boldsymbol{D}$)} through the aid of a relay satellite, in the presence of {an active} eavesdropper {at} each hop. {Moreover}, another SU located in the first hop is acting as a friendly jammer and continuously broadcasting an artificial noise that cannot be decoded by the wiretapper so as to impinge positively on the system's secrecy. Owing to the underlying strategy, the SUs are permanently adjusting their transmit powers in order to avoid causing harmful interference to primary users. The RF channels undergo shadowed-Rician and Rayleigh fading models, while the optical link is subject to Gamma-Gamma {turbulence with pointing error}. Closed-form and asymptotic expressions for the intercept probability (IP) are derived considering two different scenarios regardless of the channel's conditions, namely (i) absence and (ii) presence of a friendly jammer. The effect of various key parameters on IP, e.g., {sources'} transmit power, {artificial noise}, maximum tolerated interference power, and fading severity parameters are examined. Precisely, we aim to answer the following question: could a friendly jammer further enhance the security of such a system even in a low SNR regime? All the derived results are corroborated by Monte Carlo simulations and new insights into the considered system's secrecy are gained.