Preserving Privacy in Time Series Data Classification by Discretization

In this paper, we propose discretization-based schemes to preserve privacy in time series data mining. Traditional research on preserving privacy in data mining focuses on time-invariant privacy issues. With the emergence of time series data mining, traditional snapshot-based privacy issues need to be extended to be multi-dimensional with the addition of time dimension . In this paper, we defined three threat models based on trust relationship between the data miner and data providers. We propose three different schemes for these three threat models. The proposed schemes are extensively evaluated against public-available time series data sets [1]. Our experiments show that proposed schemes can preserve privacy with cost of reduction in mining accuracy. For most data sets, proposed schemes can achieve low privacy leakage with slight reduction in classification accuracy. We also studied effect of parameters of proposed schemes in this paper.