Using supplementary cementitious materials (SCM) can help increase the sulfate resistance of cement blends. However, formulating sulfate-resistant materials with increasing amounts of SCM is challenging, and the required standard tests last several months. Therefore, creating new tools that can be easily applied and understandable could help develop novel materials in the future. Machine learning techniques have been widely used recently to predict cementitious materials' properties such as strength, creep, or shrinkage. However, their usage is relatively limited regarding durability properties, maybe because of the large number of parameters involved in durability processes, some of them intrinsic to the material and others related to the environment. In this study, an extensive database has been built using more than 300 cementitious sample characteristics from different studies. A large collection of inputs related to cement composition, mix composition, sample geometry, and environmental conditions such as sulfate concentration has been gathered. Then several machine learning algorithms were applied to assess the resistance of blended cements to the external sulfate attack. Two groups of algorithms, e.g., classification and Regression algorithms, incorporating several models from linear to ensemble models, have then been compared. The results show that most classification models can very quickly assess the sulfate resistance of cementitious materials using the extensive database, and the best Regression models can efficiently predict the temporal evolution of the degradation. The most influential parameters can be identified, and recommendations can be drawn regarding future blended cement compositions.