Intelligent reflecting surface (IRS) is a promising technology which can be integrated with non-orthogonal mul-tiple access (NOMA) to improve the secrecy performance. In this paper, we propose an effective IRS-aided secure scheme for NOMA networks against both internal and external eavesdrop-ping. By exploiting artificial jamming (AJ), the transmit power minimization problem of legitimate signals is investigated with both users' QoS demands satisfied via the joint optimization of active and passive beamforming. The non-convex problem is decomposed into two subproblems, which are approximated into convex ones via the semidefinite relaxation (SDR). Then, by means of alternating optimization, the suboptimal solution to the original problem can be obtained. Simulation results demonstrate the superiority of the proposed scheme by combining IRS and AJ against the challenging internal and external eavesdropping.