Privacy Preserving Location Estimation

点击翻译按钮获取中文摘要

This research will investigate privacy in location-based services. The first step in providing a location-based service is to locate the user. This might compromise the user privacy. In this research, we will be developing methods in which the location of user is hidden from the location server, and the service provided is also unknown to the server. We are interested in developing location-based services where the location server cannot learn user’s location, but can compose location-related services, while not being able to identify the services obtained by the user. Also, the user can only receive services pertaining to its location, without learning about the whole space of services. In the first view this objective might be a non-starter, as the location information is needed to build the service. However, recent advances in cryptography show that certain signal processing arithmetic can be performed in the encrypted domain. In this research, we will use two methods for encrypted signal processing. As the first method, we will use additive homomorphic cryptosystems. In additive homomorphic cryptosystems, the encrypted addition of two numbers can be calculated by the multiplication of their encrypted values, and the multiplication of two numbers can be calculated by taking the encrypted value of one of them to the power of the other number. Hence, a new algebra is defined in which addition is replaced by multiplication and multiplication is replaced with the power operator. Some other operations such as Euclidean distance measurement and minimization can also be performed at the encrypted domain. We will investigate private location estimation using encrypted signal processing methods. A second group of methods for privacy preserving use randomized embedding, in which data is projected onto a lower dimensional subspace while the distance between adjacent vectors remains constant. Recently, the area of randomized embedding is receiving a lot of attention from compressive sensing and big data community. Lower dimensional embedding can hide the content of data in the higher-dimensional domain. Examples of such mappings are locally sensitive hashing, and locally linear embedding. In this research, we will build localization methods that will provide privacy preserving location estimation using both additive homomorphic cryptosystems and randomized embedding. We will develop our algorithms for client-server applications, and also for multi-party cooperative localization.