BlackSwan is an engine using sensor information from smart devices (phones, watches, fitness bands) to assist the user in his behaviour for personal security and fitness. The core element for a first application is a personal e-Call system that helps the user in a personal emergency situation (accident, robbery, heart attack etc). Further applications are e.g. Automated Assistance Systems for elder people who need assistance and Personal Fitness Monitor. The important difference to existing solutions is the functionality of the engine. Existing solutions are using static machine learning approaches, e.g. based on historical data, assumptions for future events are extrapolated. The limit (and also the complexity) are always the availability of and the access to historical data. In particular for behaviour related data, the models are coming to a natural limit, especially when it comes to applications on a smart mobile device. The BlackSwan engine is based on predictive analytics using dynamic self-adaptive and self-learning models describing normal behaviour based on collected sensor data. The strength of the system is to identify if the behaviour signals a problem for the user or if the behaviour is within a normal bandwidth. The BlackSwan engine does not require that the system already knows that such data are related to a problem for the user (like classical machine learning does). As the process does not require complex and large historical data, it can easily run on the smart device itself.