The paper discusses the challenges of modeling knowledge and communication in distributed systems with fully byzantine agents, where faulty agents can lie and deviate from protocols.
The standard epistemic analysis of distributed systems, based on the runs and systems framework and the logic S5 for knowledge, does not easily translate to byzantine settings. The authors propose a new framework that decouples the arbitrary actions of faulty agents from their knowledge, allowing agents to have false memories while retaining the logical properties of knowledge.
The authors show that knowledge is too strong a notion to trigger actions in byzantine environments, as agents cannot know that any event has objectively occurred. Instead, the authors introduce the belief modality as knowledge relativized to the agent's correctness.
To model how agents learn from communication in byzantine settings, the authors propose the hope modality. Hope represents the informational content of a message, accounting for the possibility that the sender may be faulty. The authors provide an axiomatization of the logic of knowledge and hope, and illustrate its use in reasoning about byzantine distributed systems.
Finally, the authors generalize the hope modality to the creed modality, which can model communication among agents of different types and communication strategies in heterogeneous distributed systems.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Roman Kuznet... at arxiv.org 05-07-2024
https://arxiv.org/pdf/2405.02606.pdfDeeper Inquiries