Impact analysis predicts and determines the parts of a software system that can be affected by changes of the system. Before or after such changes are made, impact analysis helps reduce the risk and costs caused by unwanted impact from changes. Traditional static impact analysis techniques, based on static system information, tend to produce imprecise results that are hardly useful. Dynamic impact analysis techniques are based on dynamic system behaviors, thus produce more precise and useful results. Existing dynamic impact analysis techniques impose various amounts of overhead costs in time and space, and produce impact sets of different degree of precision. In this paper, we propose a new dynamic impact analysis technique that is less expensive in both time and space than existing techniques, and produce safe and precise impact set relative to the dynamic information used in calculation.