In this paper, we propose a novel method to effectively detect GNSS (Global Navigation Satellite Systems) spoofing signals. Our approach utilizes mixtures of Gaussian distributions to model the Carrier Phase’s Double Difference (DD) produced by two separated receivers. DD calculation eliminates measurement errors such as ionosphere error, tropospheric error and clock bias. DD values contain the angle of arrival (AOA) information and a small amount of Gaussian noise. The authentic GNSS signals come from different directions, therefore AOA values are different for each satellite. In contrast, spoofing signals from one broadcaster should always have the same direction. Therefore, DD values of authentic satellites contain mainly the double difference of AOA values, while DD of spoofing satellites contains only an insignificant amount of Gaussian noise. That rough observation is the theoretical basis for our proposal in which we use Gaussian Mixture Models (GMMs) to learn the distribution of DD values calculated for both kinds of satellites. The pre-trained GMMs are then utilized for detecting non-authentic signals coming from spoofing satellites.