We study the exploitation of polarimetric diversity in passive multistatic radar for detecting moving targets. We first derive a data model that takes into account polarization and anisotropy of targets inherent in multistatic configurations. Unlike conventional isotropic models in which targets are modeled as a collection of uniform spheres, we model targets as a collection of dipole antennas with unknown directions. We consider a multistatic configuration in which each receiver is equipped with a pair of orthogonally polarized antennas, one directed to a scene of interest collecting target-path signal and another one having a direct line-of-sight to a transmitter-of-opportunity collecting direct-path signal. We formulate the detection of moving target problem in a generalized likelihood ratio test framework under the assumption that direct-path signal is available. We show that the result can be reduced to the case in which the direct-path signal is absent. We present a method for estimating the dipole moments of targets. Extensive numerical simulations show the performance of both the detection and the dipole estimation tasks with and without polarimetric diversity.