Various indexes are being used today to evaluate the strength of passwords. In these indexes, the strength of a password is evaluated to be high if it takes longer for an attacker to predict it. Therefore, using such an evaluation, there is a problem that a leaked password may reduce the reliability of the index by increasing vulnerability if an attacker attempts to attack using a leaked password. Hence, estimating the leaked frequency when considering strength is important for reducing vulnerability. This paper proposes a password strength evaluation model using deep learning-based multi-class classification, which solves the existing problem of leaked frequency not being considered during evaluation. Data preprocessing modeling is critical to improve the performance of this model. Additionally, since selecting and extracting feature values of preprocessing data is also important, a model that accurately estimates the degree of leakage through an evaluation method of existing indexes is proposed. To evaluate the performance of the proposed model, an experiment that compares the password leaked frequency stored in a database using a password list was conducted. As a result, the proposed model correctly evaluated 99% of the 345 leaked passwords. Therefore, the effectiveness of the proposed model was verified.