We present a method of using face color in RGB color space for discrimination between real and synthetic human faces. This involves a high-dimensional feature space, so feature dimension reduction is accomplished by the traditional method of median filtering and downsampling, and feature extraction is done with singular-value decomposition. A nonlinear support-vector machine is then used to determine whether the extracted features represent a real human face or a synthetic one. The results from our method are compared with those of traditional face recognition algorithms and are shown to have a higher rate of success even when a only small number of features are is used. The method can achieve a high level of accuracy with little computation time, and is less sensitive to facial expression variations and varying brightness levels than other methods.