A Trust–Reputation Model Offering Data Retrievability and Correctness in Distributed Storages

Abstract

Distributed storage systems employing dispersal techniques provide retrieval services with a minimal subset of available servers’ file shares. Generally, cryptography-based techniques aid data retrievability and storage correctness in distributed storages; we propose an equivalent, alternative approach that uses “server trust”. Our work makes available, file retrieval and dynamic restoration of corrupted file shares with highest probability. We present a novel trust–reputation model to quantify and manage the trust of servers operating in an untrusted environment; a significant aspect is in quantification of good/bad behaviour of servers in terms of trust index, using analytical expressions. The scheme provides selection of the most reliable, minimal server set for file reconstruction and identification of servers for restoration of corrupted file shares. Singular treatment of incorrect servers, statistical means of server reward/punishment and selection of the minimal set of file shares for reconstruction are the key features of the model. Our scheme achieves the triple goals of availability, retrievability and recoverability, despite non-crash and crash-stop faults in storage. The model has evolved into a comprehensive system satisfying the triple goals, standard framework and crucial recommendations demanded of such a system. Extensive simulation studies carried out validate the model and its adaptability.