Privacy aware publishing of successive location information in sensor networks

With the pervasive penetration of the sensor networks into people's daily life, data are becoming easily obtainable. While the information is useful in many aspects, personal privacy is greatly challenged too. In this paper, we are interested in the applications where the sensor networks are deployed to monitor the locations of a person (or an animal). While the location information is useful for the interested public or scientists, we found that a detailed knowledge of the past behavior and current track of the person can disclose his future locations; which may bring in privacy or security concerns. We call this a successive privacy problem. Notice that this is in sharp contrast to previous location privacy studies which tries to mask, through K-anonymity, an individual past or current location of a person. To date, given a sequence of past observations, abundant techniques are available to infer future locations of an object. We observe that intrinsically, each observation will contribute to the inference accuracy. Therefore, in this paper, we generalize it into a weighted representation. That is, the observations are associated with weights which show the (joint) impact on releasing the observations to inference of future data. We observed that there is an intrinsic trade-off between the number of data to be published to the interested parties and the privacy preservation of the object. We show that the problem can be formulated into a non-linear optimization problem. As the problem is intractable, we develop optimal solutions to some special cases through dynamic programming and several heuristics for the general case. We then show several privacy aware data collection schemes; their performance and efficiency. Extensive simulations demonstrate the effectiveness of our schemes.