This document describes a behavior-based authentication system for smartphones using life-log data. It collects data from smartphone sensors over 2 months from 47 students. It extracts features related to user behavior and environment, selects relevant features using random forest, builds a user behavior model based on probability distributions of daily behaviors, and measures similarity between models using Mahalanobis distance. Experimental results using data from 37 students over 6 weeks achieved an equal error rate of 7.05%, indicating good post-login security from the behavior-based authentication system.