This document summarizes research on using mobile sensor data and behavioral biometrics for user authentication and activity recognition. It describes collecting data from accelerometers, GPS, WiFi and applications to build language models of user behavior. Scores are calculated to determine the likelihood a behavior belongs to a user or activity class. Authentication is triggered based on thresholds. The system was tested to identify users from single key presses and detect anomalies with days of training data at 80% accuracy. Future work involves expanded data collection, improved models, integration with security frameworks, and ensuring user privacy.