This paper introduces a method for discovering and redacting private information in online course discussion forums to facilitate data analysis while preserving student privacy. The method involves set operations to identify potential private information such as names, nicknames, employers, hometowns, and contact details, which are then confirmed through manual annotation or machine learning. The method was tested by having two raters manually annotate a corpus of words from an online course's forum. An ensemble machine learning model was trained to automate this task, achieving high performance metrics (95.4% recall and 0.979 AUC on a same-course dataset, and 97.0% recall and 0.956 AUC on a different-course dataset). This approach was motivated by the need to answer research questions about student interactions with online courses, which required anonymized forum data. The paper also discusses perspectives from two online course instructors on the method and potential additional applications.