Dissertations details

  • Privacy-preserving efficient searchable encryption
  • Dec 2016
  • Data storage and computation outsourcing to third-party managed data centers, in environments such as Cloud Computing, is increasingly being adopted by individuals, organizations, and governments. However, as cloud-based outsourcing models expand to society-critical data and services, the lack of effective and independent control over security and privacy conditions in such settings presents significant challenges. An interesting solution to these issues is to perform computations on encrypted data, directly in the outsourcing servers. Such an approach benefits from not requiring major data transfers and decryptions, increasing performance and scalability of operations. Searching operations, an important application case when cloud-backed repositories increase in number and size, are good examples where security, efficiency, and precision are relevant requisites. Yet existing proposals for searching encrypted data are still limited from multiple perspectives, including usability, query expressiveness, and client-side performance and scalability. This thesis focuses on the design and evaluation of mechanisms for searching encrypted data with improved efficiency, scalability, and usability. There are two particular concerns addressed in the thesis: on one hand, the thesis aims at supporting multiple media formats, especially text, images, and multimodal data (i.e. data with multiple media formats simultaneously); on the other hand the thesis addresses client-side overhead, and how it can be minimized in order to support client applications executing in both high-performance desktop devices and resource-constrained mobile devices. From the research performed to address these issues, three core contributions were developed and are presented in the thesis: (i) CloudCryptoSearch, a middleware system for storing and searching text documents with privacy guarantees, while supporting multiple modes of deployment (user device, local proxy, or computational cloud) and exploring different tradeoffs between security, usability, and performance; (ii) a novel framework for efficiently searching encrypted images based on IES-CBIR, an Image Encryption Scheme with Content-Based Image Retrieval properties that we also propose and evaluate; (iii) MIE, a Multimodal Indexable Encryption distributed middleware that allows storing, sharing, and searching encrypted multimodal data while minimizing client-side overhead and supporting both desktop and mobile devices.
  • Departamento de Informática, FCT/UNL
  • Bernardo Ferreira
  • Henrique João Domingos
  • https://run.unl.pt/handle/10362/19797