Preserving Privacy in Time Series Data Mining

Time series data mining poses new challenges to privacy. Through extensive experiments, the authors find that existing privacy-preserving techniques such as aggregation and adding random noise are insufficient due to privacy attacks such as data flow separation attack. This paper also presents a general model for publishing and mining time series data and its privacy issues. Based on the model, a spectrum of privacy preserving methods is proposed. For each method, effects on classification accuracy, aggregation error, and privacy leak are studied. Experiments are conducted to evaluate the performance of the methods. The results show that the methods can effectively preserve privacy without losing much classification accuracy and within a specified limit of aggregation error.