We propose joint design on precoding and scheduling for secure device-to-device networks, wherein transmitters (TXs) send confidential messages to associated receivers (RXs) while the eavesdroppers seek to wiretap the messages. To this end, we solve a network sum secrecy spectral efficiency (SE) maximization problem. The main challenges are: i) the problem is non-convex and ii) the secrecy SE is not smooth with the presence of multiple eavesdroppers. To address the difficulties, we first approximate the secrecy SE by exploiting a smooth maximum function and transform the original problem into a tractable non-convex form. Subsequently, we propose a low-complexity algorithm to find a sub-optimal solution, referred as SecureLinQ. Simulation results validate the performance of SecureLinQ.