Privacy-preserving key-value store

• Main Author: Marques, João Bernardo Coimbra
• Publication Date: 2020
• Format: Master thesis
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10362/113608
• topic: Cloud Computing; Key-Value Stores; CRDT; Privacy; Homomorphic Encryption; Searchable Encryption.; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

Cloud computing is arguably the foremost delivery platform for data storage and data processing. It turned computing into a utility based service that provides consumers and enterprises with on-demand access to computing resources. Although advantageous, there is an inherent lack of control over the hardware in the cloud computing model, this may constitute an increased privacy and security risk. Multiple encrypted database systems have emerged in recent years, they provide the functionality of regular databases but without compromising data confidentiality. These systems leverage novel encryption schemes such as homomorphic and searchable encryp tion. However, many of these proposals focus on extending existing centralized systems that are very difficult to scale, and offer poor performance in geo-replicated scenarios. We propose a scalable, highly available, and geo-replicated privacy-preserving key value store. A system that provides its users with secure data types meant to be replicated, along with a rich query interface with configurable privacy that enables one to issue secure and somewhat complex queries. We accompany our proposal with an implementation of a privacy-preserving client library for AntidoteDB, a geo-replicated key-value store. We also extend the AntidoteDB’s query language interface by adding support for secure SQL-like queries with configurable privacy. Experimental evaluations show that our proposals offer a feasible solution to practical applications that wish to improve their privacy and confidentiality.