Query and data security in the data outsourcing model

An increasing number of applications adopts the model of outsourcing data to a third party for query and storage. Under this model, database owners delegate their database management needs to third party data servers. These servers possess both the necessary resources, in terms of computation, communication and storage, and the required expertise to provide efficient and effective data management and query functionalities to data and information consumers. However, as servers might be untrusted or can be compromised, query and data security issues must be addressed in order for this model to become more practical. An important challenge in this realm is to enable the client to authenticate the query results returned from a server. To authenticate is to verify that the query results are correctly computed from the same data as the data owner has published and all results have been honestly returned. Existing solutions for this problem concentrate mostly on static, relational data scenarios and are based on idealistic properties for certain cryptographic primitives, looking at the problem mostly from a theoretical perspective. This thesis proposes practical and efficient solutions that address the above challenge for both relational and streaming data. Specifically, this dissertation provides dynamic authenticated index structures for authenticating range and aggregation queries in both one and multiple dimensional spaces. The authentication of sliding window queries over data streams is then discussed to support data streaming applications. We also study the problem of query execution assurance over data streams where the data owner and the client are the same entity. A probabilistic verification algorithm is presented that has minimal storage and update costs and achieves a failure probability of at most δ, for any small δ 0. The algorithm is generalized to handle the scenarios of load shedding and multiple queries. An extensive experimental evaluation for all the proposed methods over both synthetic and real data sets is presented. The findings of this evaluation demonstrate both the correctness and effectiveness of the proposed methods.