The issue with data leakage can be handled through k-fold cross-validation, in which all of the data takes turns as either training data or test data.