Abstract: This paper proposes a (γ,l-p,k)-anonymity model. In the anonymous model, information entropy was introduced as the attribute distance to calculate the missing data, and the sensitive attribute degree was divided into different sensitive levels, and the corresponding weight ω value was set. At the same time, we let the number of different sensitivity levels of equivalence class tuples meet the threshold γ. According to the model, a weighted information entropy anonymity algorithm WISA* was designed to anonymize the defective dataset. The experimental results show that the proposed algorithm WISA* not only reduces the loss of equivalent information, but also improves the performance of sensitive attributes on diversity, thereby descending the risk of privacy leakage of the data with a low complexity.