The invention provides a privacy protection method based on dynamic graph data release. The privacy protection method comprises the following steps: extracting graph attribute sequences at a previousmoment and a current moment by adopting a sequence extraction generator; inputting the extracted two attribute sequences into an activation function, and calculating the value of the activation function; when the value of the activation function is smaller than or equal to a certain threshold value, performing anonymization operation on the graph by adopting a local anonymization function interface to obtain an anonymized graph; and when the value of the activation function is greater than a certain threshold, performing anonymization operation on the graph by adopting a global anonymization function interface to obtain an anonymized graph. According to the method, the activation function is adopted to determine which moment the anonymization process needs to be restarted, so that the anonymization strategy can be adaptively selected in the anonymization process, the accumulated information loss in the dynamic graph data release process is reduced, and the availability of the anonymized data is improved.