This paper proposes a method of detecting faults in non-redundant sensors. Such a method is advantageous for small unmanned aerial vehicles (UAVs), which are prevented from carrying redundant sensors due to size, weight, and power constraints. The method we propose uses a multiplicative extended Kalman filter (MEKF) for estimation and employs hypothesis testing to detect faults. This method has been shown to detect bias, drift, and increased noise in a non-redundant sensor real-time on board an autonomous rotorcraft.