A multiparty quantum secret sharing protocol based on GHZ states was proposed by Hwang et al. (Phys. Scr. 83:045004, 2011). Its major advantage is high efficiency, but soon another paper (Liu and Pan in Phys. Scr. 84:045015, 2011) shown that the protocol is insecure for one dishonest agent and give a improvement of protocol. Here, we analyze the security of both protocols, and point out that both protocols would lead to all the secret information leakage under a special attack, which is different from the attack strategy proposed by the paper (Liu and Pan in Phys. Scr. 84:045015, 2011). Furthermore, we discuss the deep reason of this insecurity, and propose a novel and efficient secure protocol, which keep the important weakness of original protocol free, i.e., correlation-extractability. The security of our scheme is equivalent to that of BB84 protocol.