Long-Term Confidential Secret Sharing-Based Distributed Storage Systems

Secret sharing-based distributed storage systems can provide long-term protection of confidentiality and integrity of stored data. This is achieved by periodically refreshing the stored shares and by checking the validity of the generated shares through additional audit data. However, in most real-life environments (e.g. companies), this type of solution is not optimal for three main reasons. Firstly, the access rules of state of the art secret sharing-based distributed storage systems do not match the hierarchical organization in place in these environments. Secondly, data owners are not supported in selecting the most suitable storage servers while first setting up the system nor in maintaining it secure in the long term. Thirdly, state of the art approaches require computationally demanding and unpractical and expensive building blocks that do not scale well. In this thesis, we mitigate the above mentioned issues and contribute to the transition from theory to more practical secret sharing-based long-term secure distributed storage systems. Firstly, we show that distributed storage systems can be based on hierarchical secret sharing schemes by providing efficient and secure algorithms, whose access rules can be adapted to the hierarchical organization of a company and its future modifications. Secondly, we introduce a decision support system that helps data owners to set up and maintain a distributed storage system. More precisely, on the one hand, we support data owners in selecting the storage servers making up the distributed storage system. We do this by providing them with scores that reflect their actual performances, here used in a broad sense and not tied to a specific metric. These are the output of a novel performance scoring mechanism based on the behavioral model of rational agents as opposed to the classical good/bad model. On the other hand, we support data owners in choosing the right secret sharing scheme parameters given the performance figures of the storage servers and guide them in updating them accordingly with the updated performance figures so as to maintain the system secure in the long term. Thirdly, we introduce efficient and affordable distributed storage systems based on a trusted execution environment that correctly outsources the data and periodically computes valid shares. This way, less information-theoretically secure channels have to be established for confidentiality guarantees and more efficient primitives are used for the integrity safeguard of the data. We present a third-party privacy-preserving mechanism that protects the integrity of data by checking the validity of the shares