Password guessing attacks and users’ vulnerable password behaviors pose severe threats to password security, and accurate password strength meters (PSMs) can mitigate these threats by nudging users to choose secure passwords. In this light, PSMs have been deployed by nearly every respectable web service and application during registration or password reset. However, most PSMs in both academia and industry only focus on the structural and semantic features of the password itself, and failed to capture users’ vulnerable password reuse behaviors. To fill this gap, we propose a new PSM, namely EditPSM, that takes users’ password reuse behaviors into consideration. EditPSM is based on a targeted password guessing framework via deep learning, and does not require users’ existing passwords to protect users’ privacy. It utilizes popular passwords identified in the training set and learns how users modify them. This leads to credential tweaking models, a type of targeted password guessing model, to effectively evaluate password strength without needing access to users’ existing passwords. Through extensive evaluations involving 10 large-scale datasets and 8 mainstream PSMs in the real world, EditPSM demonstrates its superior performance over prior art. We believe this work makes a substantial step towards introducing targeted password models into password strength evaluation.