Date on Master's Thesis/Doctoral Dissertation
Computer Engineering and Computer Science
Computer networks--Security measures; Online social networks--Security measures; Data encryption (Computer science)
Smartphones, loaded with users’ personal information, are a primary computing device for many. Advent of 4G networks, IPV6 and increased number of subscribers to these has triggered a host of application developers to develop softwares that are easy to install on the mobile devices. During the application download process, users accept the terms and conditions that permit revelation of private information. The free application markets are sustainable as the revenue model for most of these service providers is through profiling of users and pushing advertisements to the users. This creates a serious threat to users privacy and hence it is important that “privacy protection mechanisms” should be in place to protect the users’ privacy. Most of the existing solutions falsify or modify the information in the service request and starve the developers of their revenue. In this dissertation, we attempt to bridge the gap by proposing a novel integrated CLOPRO framework (Context Cloaking Privacy Protection) that achieves Identity privacy, Context privacy and Query privacy without depriving the service provider of sustainable revenue made from the CAPPA (Context Aware Privacy Preserving Advertising). Each service request has three parameters: identity, context and actual query. The CLOPRO framework reduces the risk of an adversary linking all of the three parameters. The main objective is to ensure that no single entity in the system has all the information about the user, the queries or the link between them, even though the user gets the desired service in a viable time frame. The proposed comprehensive framework for privacy protecting, does not require the user to use a modified OS or the service provider to modify the way an application developer designs and deploys the application and at the same time protecting the revenue model of the service provider. The system consists of two non-colluding servers, one to process the location coordinates (Location server) and the other to process the original query (Query server). This approach makes several inherent algorithmic and research contributions. First, we have proposed a formal definition of privacy and the attack. We identified and formalized that the privacy is protected if the transformation functions used are non-invertible. Second, we propose use of clustering of every component of the service request to provide anonymity to the user. We use a unique encrypted identity for every service request and a unique id for each cluster of users that ensures Identity privacy. We have designed a Split Clustering Anonymization Algorithms (SCAA) that consists of two algorithms Location Anonymization Algorithm (LAA) and Query Anonymization Algorithm (QAA). The application of LAA replaces the actual location for the users in the cluster with the centroid of the location coordinates of all users in that cluster to achieve Location privacy. The time of initiation of the query is not a part of the message string to the service provider although it is used for identifying the timed out requests. Thus, Context privacy is achieved. To ensure the Query privacy, the generic queries (created using QAA) are used that cover the set of possible queries, based on the feature variations between the queries. The proposed CLOPRO framework associates the ads/coupons relevant to the generic query and the location of the users and they are sent to the user along with the result without revealing the actual user, the initiation time of query or the location and the query, of the user to the service provider. Lastly, we introduce the use of caching in query processing to improve the response time in case of repetitive queries. The Query processing server caches the query result. We have used multiple approaches to prove that privacy is preserved in CLOPRO system. We have demonstrated using the properties of the transformation functions and also using graph theoretic approaches that the user’s Identity, Context and Query is protected against the curious but honest adversary attack, fake query and also replay attacks with the use of CLOPRO framework. The proposed system not only provides 'k' anonymity, but also satisfies the < k; s > and < k; T > anonymity properties required for privacy protection. The complexity of our proposed algorithm is O(n).
Pandit, Anala Aniruddha, "Privacy protection in context aware systems." (2014). Electronic Theses and Dissertations. Paper 1092.