º£½ÇÉçÇø

[Defense] Enhancing DAG-Based Byzantine Fault Tolerance with augmented Graph Neural Networks for Scalable and Adaptive Blockchain Consensus

Wednesday, December 4, 2024

3:00 pm - 4:30 pm

In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
Nour Diallo

will defend his proposal
ÌýEnhancing DAG-Based Byzantine Fault Tolerance with augmented Graph Neural Networks for Scalable and Adaptive Blockchain Consensus

Abstract

In this proposal, we introduce DAGWise, a novel system that integrates Graph Neural Networks (GNNs) with Directed Acyclic Graph-based Byzantine Fault Tolerant (DAG-BFT) consensus protocols to address key challenges in distributed consensus. By embedding GNNs into the core of the DAG-BFT architecture, DAGWise dynamically optimizes leader selection and message validation processes, which are traditionally susceptible to inefficiencies and vulnerabilities under fluctuating network conditions. This adaptive system not only enhances decision-making accuracy but also proactively identifies and mitigates potential bottlenecks and malicious activities, a capability that is essential in maintaining the robustness and security of distributed networks. DAGWise utilizes GNNs to analyze complex inter-node relationships, enabling predictive modeling of network behaviors and contributing to timely adjustments in the consensus process. This predictive capability allows the system to refine message pipelining and validation pathways, which significantly reduces communication overhead and ensures swift response times. The architectural design of DAGWise also incorporates advanced mechanisms for load balancing and fault tolerance, enhancing scalability in environments with variable network loads. Performance benchmarks indicate that DAGWise delivers substantial improvements in throughput and latency over conventional DAG-BFT systems, establishing it as a highly resilient solution for modern distributed systems. Furthermore, our approach underscores the potential of GNNs in enhancing the adaptability and security of consensus protocols, highlighting DAGWise as a viable model for future applications in high-stakes, dynamic distributed networks. Through this research, we aim to contribute to the field of consensus mechanisms by offering a robust framework that can accommodate the growing demands of decentralized applications while preserving efficiency and security across diverse operational conditions.

Wednesday, December 4, 2024
3:00 PM - 4:30 PM

PGH 591

Dr. Weidong (Larry) Shi, proposal advisor

Faculty, students, and the general public are invited.