LBS Privacy Protection Based On Dynamically Hiding Sensitive Association Rules

With the development of mobile communication and computing technology, location-based services(LBS) have been developed and widely used. In short, LBS are value-added services that provided by service provider based on user’s location, such as navigation services, location-based query services, etc.. While LBS providing convenient service for people, the mobile user’s personal information leakage has gradually become a serious problem, and caused a wide range of related research. Privacy protection has become a key issue to be solved in the further development of LBS.In the traditional LBS privacy protection technology, the spatial-temporal k-anonymous method is simple and flexible and more suitable for mobile computing environment, is the most widely used model. Traditional privacy protection method based on temporal and spatial-temporal k-anonymity is difficult to effectively deal with the inference attack on sensitive association rule mined from anonymity sets. This paper mainly consider the protection of inference attacks of sensitive association rules mined from anonymity set. According to the research of privacy preserving data mining, based on process of border revision which is a classical knowledge hiding method, and knowing that this static method is not applicable for LBS, the paper proposes an improved border revision method based on the extension of transaction set. Without reconstructing the original data, the method is able to hide the sensitive association rules by extending the original transaction set. Appling the transaction set extension method to the spatial-temporal k-anonymous, sensitive association rules can be hidden through the extension of anonymity sets in order to protect location privacy. Experiments show that the method can effectively hide the sensitive association rules mined from anonymous sets and avoid the creation of the new association rules and the loss of original rules, which means the method has minimal side effect. In addition, the method is self-adapting for the dynamic update of the anonymity sets and sensitive association rules, which is in accordance with the characteristics of long-term, dynamic and online of LBS.