Password recovery of WPA2-PSK is an important problem in digital forensics. Since the encryption mechanism of WPA-PSK is gradually enhanced, it is difficult to deal with this problem by the traditional methods such as brute force, rainbow table, Markov model, and so on. In this paper, we give a new method based on simulated annealing (SA) and hidden markov model (HMM). The main principle of this method is to create the hidden markov model of the known password based on the SA which could be used to generate the password candidates in the wireless network password recovery. It means that the passwords are given by a probability learning of the known password. The tests have shown that this approach could improve the effectiveness of password recovery for the wireless network, comparing with the Markov model which has been shown much more efficiently than the traditional methods such as brute force and dictionary attack.