Hashgraph Over the Edge: Achieving Byzantine Fault Tolerance in Decentralized P2P Frameworks

This paper explores the integration of the Hashgraph consensus protocol into two widely used peer-to-peer (P2P) communication frameworks: Secure Scuttlebutt (SSB) and libp2p. SSB operates within a local-first architecture, prioritizing offline-first data synchronization through append-only logs and gossip-based replication, while libp2p employs a modular, pub-sub-driven approach to facilitate real-time message dissemination across decentralized networks. We propose a dual integration approach, embedding an asynchronous Byzantine fault-tolerant (aBFT) layer into both frameworks to evaluate their complementary strengths in diverse networking scenarios. This dual-path strategy allows us to investigate how Hashgraph’s aBFT properties interact with fundamentally different P2P architectures—append-only, local-first versus modular, pub-sub—thus offering a broader understanding of its applicability. Our goal is to enhance their ability to achieve a globally consistent state <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">despite the challenges of intermittent connectivity and network partitions</i>. To assess the feasibility and performance implications of this integration, we conduct a series of controlled experiments using Docker-based network simulations. Our evaluation focuses on key trade-offs involving consistency, latency, and bandwidth utilization. The results show that SSB+Hashgraph achieves strong eventual consistency, ensuring robust data integrity even under adversarial conditions. However, this comes at the cost of increased message propagation latency, due to the bulk gossip overhead inherent in SSB’s replication model. In contrast, libp2p+Hashgraph enables significantly faster message propagation through its pub-sub mechanism, but incurs higher bandwidth consumption as more frequent state updates are required to maintain consensus. Notably, our experiments confirm that the proposed approach successfully maintains consensus even in the presence of up to 25% malicious nodes, highlighting its resilience and strong Byzantine fault tolerance. These findings demonstrate the practical viability of combining gossip-based data replication with aBFT consensus to improve the reliability and scalability of decentralized applications. Moreover, our study offers valuable insights into the trade-offs involved in designing resilient P2P communication systems capable of operating efficiently across heterogeneous network environments.