Research on Personal Evaluation Data Collection and Protection Based on Differential Privacy
Yue Wu
State Grid Gansu Electric Power Company, Lanzhou 730030, Gansu Province, China
Yaping Pan
State Grid Gansu Electric Power Company, Lanzhou 730030, Gansu Province, China
Gang Wang
Gansu Tongxing Intelligent Technology Development Co., LTD., Lanzhou 730050, Gansu Province, China
Shenghong Wang
State Grid Gansu Electric Power Company, Lanzhou 730030, Gansu Province, China
Zhenfen Zhang
Gansu Tongxing Intelligent Technology Development Co., LTD., Lanzhou 730050, Gansu Province, China
Abstract:
Personnel evaluation data involved in decision-making and feedback, mostly in the form of collections, the data structure is complex, if the direct collection and analysis of these data will bring serious privacy and security issues, but the existing privacy protection mechanisms for collection-type data if directly applied to the personnel evaluation data transmission space will have a great impact on the accuracy of the frequency distribution estimation results, affecting the overall effect of the data transmission, and reducing the data transmission privacy. Therefore, a personnel evaluation data collection and protection method based on differential privacy is proposed. Through the description of personnel evaluation data collection and protection problems, the local differential privacy data collection and protection scenario for personnel evaluation data is constructed. Based on the local differential privacy method, the personnel evaluation data processing architecture is established, the user-side data privacy collection taking into account random perturbation and the personnel evaluation data protection considering the server-side distribution estimation are carried out, and the evaluation data protection is realized by restoring the real personnel evaluation data according to the frequency distribution. The experimental results show that in the process of privacy processing for evaluation data, the highest value of data leakage possibility obtained by utilizing the design method is 0.315, and the overall complexity value of data structure is 16.23%. It shows that the utilization of the design side effectively reduces the possibility of privacy data leakage as well as the complexity of the data results itself, and has a better protection effect on the data.