Medical data release anonymization method

The invention provides a medical data release anonymization method. The method comprises the following steps of: constructing (w, k, d) an anonymous model, giving different sensitivity scores by combining different disease sensitivities with an actual evaluation standard, dividing sensitive attributes into different levels by using the scores, calculating a weight value of each disease, and limiting an average weight value in each equivalence class to be less than a given constraint value w0; and secondly, constraining the distance of the diseases from the semantic level, ensuring the semantic diversity, requiring that the average distance in the equivalence class is greater than a given constraint value d0, and finally realizing personalized protection of sensitive attributes. According to the model, leakage of the implicit attribute of the serious disease is protected from the dimension of the sensitivity level, leakage of the implicit attribute of the disease part or organ is also protected from the dimens