Body sensor networks (BSNs) are an emerging class of medical cyber-physical systems which have the potential to change the healthcare paradigm. However, they present many new challenges, chief of which (like any medical device) is assuring patient safety. This requires not only a precise definition of safety, but also techniques for assessing the safety of BSN designs. Although solutions are possible and important for specific BSNs used in specific applications, addressing this issue on a case-by-case basis usually results in an ad-hoc process, and more importantly, makes the translation of experiences and solutions between different applications more difficult.A generic and conceptual framework for guiding the safety analysis process would provide all stakeholders a common basis for communicating, discussing, and examining the safety of BSN designs, and provide manufacturers with an exemplary process that they can follow to improve and gain confidence in the safety of their devices. This paper presents our current efforts in developing such a framework. In particular, we present a theoretical foundation for modeling and analyzing BSNs, and identify the general class of hazards based on this foundation. These efforts explore critical issues that deserve attention in designing safe BSN systems, and more importantly, can help advance the understanding of BSNs and their safety.