State Machine Replication (SMR) is a classical approach to implement consistent fault-tolerant services. This approach can also be used to implement intrusion tolerant services – which maintain integrity and availability in the presence of Byzantine faults. However, in order to guarantee confidentiality, a proactive secret sharing scheme must be used to split data into shares to be distributed between servers. These shares should be periodically renewed to protect the secrecy of the data against a mobile adversary. Current share renewing protocols are designed to renew shares of a single secret , and thus are very inefficient when applied to systems that stored a large number of shares. We propose a genetic framework called COBRA that successfully tackles such scalability problem. COBRA integrates Feldman’s secret sharing scheme to split data into data into shares with Herzberg et al. proactive scheme to periodically renew shares and allow the recovery of a server’s state thus providing protection against a mobile a adversary. Furthermore, the framework allows the framework allows the reconfiguration of the system at runtime. We mitigated the impact of recovering a large state composed of multiple shares by proposing an Optimistc Polynomial interpolation (OPI) scheme that integrates Harn and Lin detection scheme to mitigate the cost of verifying feldman’s commitments in faul-free scenarios.We implemented COBRA on top of BFT-SMaRT and evaluated it to undersand the impact of integrating secret sharing into BFT SMR. As expected, the results show that integration of secret sharing layer has a negative impact on the system. However, by emplyoing the OPI scheme, we significantly improved the latency of a share recovery and renew protocls.

Tese de mestrado, Engenharia Informática (Arquitetura, Sistemas e Redes de Computadores) Universidade de Lisboa, Faculdade de Ciências, 2019