Cloud computing is one of the major growth areas in ICT and will be at the heart of cyber physical systems and smart applications of the future. Although not being a fundamental new technology from a computer science perspective, it is disruptively changing the provisioning models for IT resources and services, i.e., the way we consume compute and storage resources as well as software solutions. Large cloud infrastructures allow for on-demand self-service of infrastructure, platform, and services at a very competitive price and on a pay per use basis. This can excel many business opportunities and helps to increase the innovation capacities abroad. However, this IT delivery model in most cases also means the adoption of an outsourcing model with all its associated problems from a security perspective. Running business critical processes on shared infrastructures without direct control introduces many new risk and demand for new protective measures in order to mitigate the respective dangers.
Data breach and data loss are two of the most prevailing risks in cloud usage and deters especially the business sector from adopting the technology. Prismacloud is dedicated to the development of enabling tools and controls which help to protect from the security risks and therefore build more trustworthy services. One part of PRISMACLOUD is dedicated to the development of protective measures for data at rest beyond simple encrypt-then-store scenarios which are only useful in single user scenarios without the need for sharing or collaboration. Additionally, naive encryption only protects the condentiality of data but introduces further risks for data loss in the case of lost or destroyed encryption keys.
In PRISMACLOUD we are aiming at the development of distributed storage systems which protect from data breach and loss at the same time. We are investigating possible solutions for storage layers based on secret sharing, a versatile cryptographic primitive ideally suited to be used as secure data dispersal algorithm. The application of threshold secret sharing allows to encode data into multiple parts such that only a pre-dened number of shares is required to recover the information. While naive application of plain secret sharing algorithms seems straight forward a more detailed analysis reveals many diculties and constraints being considered and addressed in cloud-of-cloud settings.
This report is dedicated to the review of available secret sharing techniques and categorizes them according their properties. We distinguish them according the condentiality properties they provide (information theoretical secure vs. computational secure), according the network model they are designed for (synchronous vs. asynchronous) as well as the protocol level characteristics they support. The report provides a comprehensive analysis of state-of-the-art secret sharing primitives and presents most ecient implementations regarding the requirements we have in PRISMACLOUD for building a secure and privacy preserving cloud based data sharing solution in a cloud-of-clouds setting.
Complementary to this document, Deliverable D5.1 [DDH+16] shows how actual distributed storage solutions can be built based on the primitives reviewed in this report.