In this paper, we consider the problem of recognizing human faces from the facial images of front views against random corruption. To this end, we develop a Huber collaborative representation based classification (HCRC) approach and apply it to face identification. To handle gross corruption, we exploit the robust Huber estimator as the cost function. A half-quadratic optimization algorithm is devised to solve the HCRC model efficiently. The experiments on real-life data validate the efficacy of HCRC for face identification.