基于KL-散度的电力用电数据自动脱敏算法研究

Research on Automatic Desensitization Algorithm of Power Consumption Data Based on KL-divergence

电力数据中包含一些隐私数据,一旦泄漏,就会对个人隐私安全造成隐患。为保证电力数据的安全性,提出了基于KL-散度设计电力用电数据自动脱敏算法。基于KL-散度算法建立敏感数据过滤模型,计算不同变量数据的KL距离,得到其相似性指标,对用户项目评分进行平滑处理,将具备相似性的敏感数据分成不同的批次。敏感数据去身份化处理,将数据匿名转换,计算用户真实路径被泄露的概率。设计数据自动脱敏算法,分别计算概念化数据、元组信息以及信息流的损失程度,以此判定脱敏后的数据是否可用。检验脱敏前后数据一致性,三类电力用电数据的变化率分别为0.43%、0.14%和0.11%,远远小于标准值。且算法在运行过程中单位时间处理数据量和平均延迟时间也较为理想,可见该脱敏算法具备实用性。

Power data contain some private data,which will cause hidden danger to personal privacy security once leaked.To ensure the security of power data,this paper proposes an automatic desensitization algorithm for power data based on KL-divergence.The sensitive data filtering model is established based on the KL-dispersion algorithm,and the KL distance of different variable data is calculated to obtain its similarity index.The user item score is smoothed,and the sensitive data with similarity is divided into different batches.The sensitive data are de-identified,the data are converted anonymously,and the probability of the user’s real path being leaked is calculated.The automatic desensitization algorithm is designed to calculate the loss degree of conceptual data,tuple information and information flow respectively to determine whether the desensitized data are available.This paper checks the data consistency before and after desensitization.The change rates of the three types of electric power consumption data are 0.43%,0.14%and 0.11%,which are far less than the standard value.In addition,the amount of data processed per unit time and the average delay time of the algorithm are also relatively ideal during the operation,which shows that the desensitization algorithm is practical.