Fast, Fair, and Secure Hierarchical Consensus and Dissemination for Blockchain Resilience despite Data-Center Disasters

Fairness, efficiency, and resilience to faults and attacks are fundamental properties of large-scale distributed systems to maintain data consistency across nodes in applications like blockchains and distributed databases. These systems rely on robust transaction dissemination and consensus mechanisms to face the rising challenges from accidental faults and sophisticated cyberattacks. To tolerate disasters, such as power outages and natural disasters, that could incapacitate entire data centers, implementations must be geo-replicated across multiple data centers.

This thesis explores fault tolerance and resilience strategies for the dissemination and consensus layers of cloud-based distributed systems. Recent high-profile incidents underscore the need for these systems to withstand both external threats, such as denial-of-service attacks, and inside threats, such as front-running and transaction reordering, where malicious actors manipulate transaction sequences for personal profit. The construction of reliable dissemination and consensus layers has proven to be complex and error-prone, motivating this work’s approach of composing secure and scalable solutions from existing protocol building blocks.

This work introduces two such protocols, Orion and Hermes: Orion is a hierarchical Byzantine fault-tolerant consensus protocol, designed for geo-replicated data centers, to enhance both scalability and disaster tolerance. A key innovation in Orion is its novel compositionality result, allowing the application of cluster confirmation to replace conventional trusted components, while upholding fault tolerance even in the presence of all Byzantine clusters. This allowed Orion to combine HotStuff for local intra-cluster consensus with the hybrid protocol Damysus for inter-cluster coordination. Orion’s compositional structure is optimized for performance, achieving a 20\% throughput improvement over hierarchical protocols like GeoBFT by exploiting parallelism while avoiding resource-intensive view-changes.

Hermes is a fair and resilient information dissemination protocol for permissioned and permissionless blockchain networks. Designed to counter transaction reordering and front-running attacks, Hermes utilizes robust overlay structures that are optimized for fair and low-latency dissemination. It protects against the above attacks by randomizing dissemination paths, making nodes accountable for the paths they have chosen. This ensures that no single node can dominate or manipulate the order of information flow. Comparative evaluations demonstrate Hermes’s superiority over traditional gossip and broadcast methods, achieving lower latency and improved network efficiency, thereby enhancing fairness and security in dynamic blockchain environments.

In combination, Orion and Hermes present a comprehensive framework fault and distaster tolerant, secure and efficient blockchain systems.