A Collision Mitigation Brake System (CMBS), which is mainly focused on rear-end collisions, was introduced in the Japanese market in June 2003. To make such kinds of advanced driver assistance systems more available in and accepted by society, it is essential to measure their effectiveness in enhancing safety. However, it is difficult to estimate the reduction in the number and severity of accidents quantitatively, because crash data rarely contain enough detail regarding the pre-crash accident scenarios. Such data are very important to predict how well such technologies can work when a collision is impending. In this study, a new approach was developed for technology effectiveness estimation using a simulation model and applying it to CMBS evaluation. The simulation model consists of the accident scenario database, the vehicle model, the driver model, and the environment model. We reconstructed accident scenarios of about 50 cases for rear-end collisions from US National Automotive Sampling System / Crashworthiness Data System data, resulting in time histories of striking and struck vehicles such as velocity, heading angle, trajectory, relative movements, and struck position. The vehicle model includes a radar model, CMBS control logic, and a brake actuator model as well as a conventional vehicle dynamics model. The driver model, which can react to the warnings of CMBS by braking and/or steering, was based on test results using a driving simulator. We first ran the simulations using the vehicle model without CMBS and calibrated the necessary parameters such as delta V with the accident data. Then CMBS was added to the system, and simulations were run repeatedly with some Monte Carlo type variations of variables such as driver's response time and amount of maneuver. Finally we estimated the probability of fatality and other injury indices based on the calculated delta Vs. The results showed that CMBS has substantial potential to reduce or mitigate rear-end collisions.