New smartphone technologies for the first time provide a platform for a new type of on-person, public health data collection and also a new type of informational public health intervention. In such interventions, it is the device via automatically collecting data relevant to the individual's health that triggers the receipt of an informational public health intervention relevant to that individual. This will enable far more targeted and personalized public health interventions than previously possible. However, furthermore, sensorbased public health data collection, combined with such informational public health interventions provides the underlying platform for a novel and powerful new form of learning public health system. In this paper we provide an architecture for such a sensor-based learning public health system, in particular one which maintains the anonymity of its individual participants, we describe its algorithm for iterative public health intervention improvement, and examine and provide an evaluation of its anonymity maintaining characteristics.