The masked data are the system failure data when the exact cause of the failures might be unknown. That is, it is the subset of components known to contain the component causing system failures. In general, the maximum likelihood estimation (MLE) of parameters are difficult to find when there exist masked data, because superposition non-homogenous Poisson process (NHPP) software reliability model cannot be decomposed into the independent NHPP models. In this paper, the MLE of software reliability from masked data is studied based on superposition NHPP models. Finally, numerical example based on simulation data is given to illustrate a good performance of MLE.