The article discusses a methodology for constructing a binary classifier to identify crypto wallets associated with ransomware. A dataset of 41,698 addresses was created, of which 20,849 were ransomware-related and 20,849 were non-ransomware. For each of the wallets, 53 features were identified. The following algorithms were used to build classifiers: logistic regression, k-nearest neighbors, decision trees, random forest, gradient boosting. The hyperparameters of the classifiers were selected. To assess the quality of the classifiers, the following metrics were used: accuracy, precision, recall, F1, ROC-AUC and ROC curves. According to the accuracy (95.54%), precision (92.40%), F1 (94.24%) metrics, gradient boosting showed the best result, according to the recall metric (99.15%) - logistic regression, according to the ROC-AUC metric (98.85%) - random forest. According to ROC curves - random forest and gradient boosting.