Privacy Preservation By K-Anonymizing Ngrams Of Time Series

Time series data, such as ECG, can be shared publicly for data mining applications and researches. This data similar to different kind of data types could be illegally exploited by an adversary to reveal identity of an individual. To prevent reidentification, many k-anonymization methods are introduced. Predictive models use probabilities of Ngrams of time series to predict future values. In this paper we propose an algorithm for k-anonymization of Ngram models of time series. It hides rare Ngrams of the time series between all other Ngrams that their frequencies are guaranteed to be at least k. Utilizing proposed algorithm on the real time series shows its effectivity by maximum information loss 2%.