Probabilistic competing failure analysis in body sensor networks

Body Sensor Networks (BSNs) have been developed to provide real-time, ubiquitous health monitoring and diagnosis for human beings. Due to their life-critical applications, it is significant to model and evaluate reliability of BSNs. A ch allenge is to consider probabilistic competing failures of different BSN components. Particularly there exists a competition in the time domain between the propagated failure of a biomedical sensor (e.g., due to jamming attack) and the local failure of a relay node which is responsible for delivering the physiological information sensed by the biomedical sensor to the sink device. Different failure orderings can lead to dramatically different BSN status. In addition, each biomedical sensor in a BSN can experience two types of local failures which have different statistical relationships with the propagated failure originating from the same sensor. This paper models the reliability of BSNs considering the two different statistical relationships between local and propagated failures of biomedical sensors as well as competing failure propagation and probabilistic failure isolation effect. A combinatorial and analytical method is proposed, which is applicable to arbitrary types of distributions for both component failures and probabilistic failure isolation factors. The proposed method is illustrated through the detailed analysis of an example BSN system.