In power distribution networks, microgrids utilize Phasor Measurement Units (PMUs), to assess the voltage stability at critical nodes in the network. PMUs rely on precise time-keeping sources, such as GPS, to obtain synchronization. However, GPS signals are vulnerable to external spoofing attacks due to their unencrypted signal structure and low received power. To detect the spoofing-induced timing anomaly, an innovative geographically Distributed Multiple Directional Antennas (DMDA) setup is proposed, which is triggered using a common clock. Utilizing the configuration of the proposed DMDA, a Belief-Propagation (BP)-based Extended Kalman Filter (EKF) algorithm is developed to estimate the timing errors caused by spoofing. The BP-EKF algorithm analyzes the single difference pseudorange residuals across each pair of antennas in a probabilistic graphical framework not only to detect the spoofed antennas in the DMDA setup but also to estimate the timing errors associated with the spoofed antennas. Based on the BP estimate of timing error at each antenna and the known baseline distances across antennas, the pseudoranges are corrected, and then adaptive EKF is employed to estimate the GPS timing. The performance of the BP-EKF algorithm is assessed by subjecting the simulated authentic GPS signals to a simulated meaconing attack, which induces a time delay of 60 microseconds. Both successful detection of meaconing, and also accurate estimation of GPS timing that complies with the IEEE-C37.118 standards, is validated using the experimental results. At a critical node in the simulated microgrid, as compared to scalar tracking, an increased voltage stability is demonstrated using the BP-EKF by assessing a metric, namely, voltage stability index.