Research on building energy systems is shifting toward data-driven modelling due to the complications of setting up physics-based models. Data-driven learning from buildings faces an important challenge, namely, accessibility to input data for training the models. The scarcity of building energy data is mainly derived from privacy concerns, which are of greater importance for residential buildings. Restrictions imposed on information prevent researchers from sharing data for reproduction and benchmarking purposes and impede comparison between different models. Therefore, there is a clear void concerning open datasets of building energy consumption, which may impair the advancement of machine learning and data mining for building energy research. One viable solution to overcome the barrier for open data sharing is through data anonymization. Recent advancements in machine learning have introduced reliable models that suitably anonymize data with little information loss and high security. However, the building research community still face a number of challenges that are vital for the safe and reliable dissemination of data in an open manner. This chapter introduces the hurdles, describes the most promising methods, and provides a synthesis of the next steps to overcome the existing barriers to open data sharing in the building research community.