A framework to overcome the effect of radar glint noise for reentry-phase tracking of a ballistic missile is presented. Glint noise is characterized by probability density functions that are heavy-tailed. For such cases, the conventional Gaussian modeling of measurement noise does not apply. A filter designed with a Gaussian noise assumption will incur a performance penalty when outliers due to glint noise are encountered. In this work, the classic Extended Kalman Filter (EKF) is modified using a robust M-estimation procedure that accounts for measurement noise distributions with heavy tails. The proposed method provides a substantial improvement (in terms of root-mean-square error) over the standard versions of EKF and UKF. All the benefits that make EKF the method of choice for state estimation (e.g. simplicity and computational efficiency) are retained in this work while robustness is added. The suggested technique is validated using simulation experiments on a three-dimensional ballistic missile reentry model.