This paper proposes a dynamic information fusion framework for sensor networks with the integration of local observers, value of information, and active-passive consensus filters as well as a layer to monitor the validity of information. Specifically, we consider a process of interest consisting of multiple subprocesses (for example, multiple targets to be monitored). The heterogeneity in the sensor networks is considered and handled in many aspects such as nodes are allowed to have different sensing capabilities, different information node roles (active and/or passive; that is, a node can be subject to observations of the process or to no observation), and different weights on information (value of information). In addition, the information validity monitor layer allows operators to evaluate the reliability of the fused information based on the local feedbacks received from the sensor network. Several illustrative numerical examples are also presented to illustrate the efficacy and discuss the practical aspects of the proposed dynamic information fusion framework.