Face recognition using visible-image (RGB) has been found to be largely affected by variation of illuminations and face expressions. Relatively, the infrared-image (IR) shows certain robustness with respect to the above external factors. In this work, we focus on identity verification using IR face images and study its robustness in response to variation of above environmental factors and expression. The well-known Principal Component Analysis (PCA) technique and another dimension reduction method, Random Projection (RP) are adopted to extract necessary IR face features for identity matching. To boost the verification performance, several fusion techniques are proposed and investigated. Our empirical experiments show reasonable effectiveness using IR and its contour images as well as using fusion method for identity verification under several varying conditions.