Vast amounts of digital footprint data available from sources like social networks and smartphones provide a wealth of information to study human behavior and social interactions at scale. Given a set of social interactions of a user, how can we detect changes in interaction patterns over time? How intrusive are these analyses to user privacy? While most previous works focus on studying network-wide properties and spotting outlier users, the dynamics of individual user interactions remain largely unexplored. We focus on the following topics: a) pro filing users at group and individual levels to spot anomalous behaviors b) protecting user privacy while maintaining prediction accuracy. For the first, we introduce statistical modeling methods for group and personalized profiling using temporal timelines of user activities. We use dynamic ego networks of user interactions to spot suspicious periods of activity. For the second, we make employ tensor decomposition methods to analyze user temporal data, exploring the trade-offs between identifying high-quality user clusters while protecting individual user privacy.