MoLe is a system that uses sensors in smartwatches to detect keystrokes by analyzing motion data during typing. It identifies keystroke-related movements using a bagged decision tree classifier and fits point clouds to determine centroids of typed characters. A Bayesian inference model incorporates sequential typing patterns and speed factors to assign probabilities to candidate words based on sensor observations. An evaluation with 8 subjects typing 300 words showed MoLe could guess words within the top 30% for 5 candidates and top 50% for 24 candidates. While sensor data leaks information, sampling rates can be reduced to mitigate these attacks. Wearables present both benefits and security risks that require consideration.