UVM Theses and Dissertations
Format:
Print
Author:
Wang, Song
Dept./Program:
Computer Science
Year:
2013
Degree:
PhD
Abstract:
At present, the widespread adoption of OPS-enabled devices has allowed a large amount of location data to be collected. This data can take the form of approxiniate location information reported by vehicles moving through road networks or estimated counts of people moving in indoor spaces such as shopping malls, office buildings or amusement parks. Itcan also be the location di~tribution of different diseases or accidents. It can even be the location of roaming taxis or people who request taxi service. In short, location data is abundant and diverse.
This data can be useful for a variety of applications, inclu~glocation-based services (e.g., Google Map and Latitude), business intelligence such as indoor shopping assistance planning, information guiding such as taxi scheduling, and so forth. However, location data usually contains sensitive information such as location and identity of the application subscriber, which may pose severe privacy threats. As a result, privacy leakage in location data may not only limit the popularity of location-based services, but may also inhibit the usage of location data in broader domains.
In this dissertation, we investigate three challenging yet fundamentally related problems regarding locationdata. Firstly, we study how to hide sensitive information in location data. Two main types of privacy have been identified, namely, identity and location privacy. In order to protect these two types of privacy, it helps to know the distribution of moving objects because the feasibility of both state-of-the-art and our proposed approaches rely on the underlying distribution of moving objects. Secondly, we study the distribution of moving objects in indoor spaces. Given the distribution of different spatial objects, an important application is to find out regions where different spatial features co-occur with a significantly higher probability inside those regions than the outside. This problem is known as spatial co-location in the research community. Thus, thirdly, we propose a statistical framework to study the problem of regional co-location with arbitrary shapes.
For the first problem, we devise two algorithms to protect the identity and location privacy of moving objects, namely, AnonTwist and Boxed algorithms. Our algorithms are efficient and effective in providing both location and identity privacy protection in terms of anonymity and privacy-safe region size. We also propose a new architecture to protect identity privacy. For the second problem, we propose a particle filter-based framework to estimate the distribution of moving objects in indoor spaces with noisy sensors. Our framework takes advantage of continuous, albeit inaccurate sensor readings. Our particle filter-based framework can estimate the distribution of moving objects in indoor spaces with high accuracy. For the third problem, we propose an efficient framework to find spatial co-location patterns with arbitrary shapes. Our method can discover the co-location patterns with high accuracy.
The algorithms proposed in this dissertation solve three important and related research problems regarding location data. Their solutions help future research in other areas including indoor data management, data management in sensor networks and spatial data mining.
This data can be useful for a variety of applications, inclu~glocation-based services (e.g., Google Map and Latitude), business intelligence such as indoor shopping assistance planning, information guiding such as taxi scheduling, and so forth. However, location data usually contains sensitive information such as location and identity of the application subscriber, which may pose severe privacy threats. As a result, privacy leakage in location data may not only limit the popularity of location-based services, but may also inhibit the usage of location data in broader domains.
In this dissertation, we investigate three challenging yet fundamentally related problems regarding locationdata. Firstly, we study how to hide sensitive information in location data. Two main types of privacy have been identified, namely, identity and location privacy. In order to protect these two types of privacy, it helps to know the distribution of moving objects because the feasibility of both state-of-the-art and our proposed approaches rely on the underlying distribution of moving objects. Secondly, we study the distribution of moving objects in indoor spaces. Given the distribution of different spatial objects, an important application is to find out regions where different spatial features co-occur with a significantly higher probability inside those regions than the outside. This problem is known as spatial co-location in the research community. Thus, thirdly, we propose a statistical framework to study the problem of regional co-location with arbitrary shapes.
For the first problem, we devise two algorithms to protect the identity and location privacy of moving objects, namely, AnonTwist and Boxed algorithms. Our algorithms are efficient and effective in providing both location and identity privacy protection in terms of anonymity and privacy-safe region size. We also propose a new architecture to protect identity privacy. For the second problem, we propose a particle filter-based framework to estimate the distribution of moving objects in indoor spaces with noisy sensors. Our framework takes advantage of continuous, albeit inaccurate sensor readings. Our particle filter-based framework can estimate the distribution of moving objects in indoor spaces with high accuracy. For the third problem, we propose an efficient framework to find spatial co-location patterns with arbitrary shapes. Our method can discover the co-location patterns with high accuracy.
The algorithms proposed in this dissertation solve three important and related research problems regarding location data. Their solutions help future research in other areas including indoor data management, data management in sensor networks and spatial data mining.