Open access
Author
Date
2020Type
- Doctoral Thesis
ETH Bibliography
yes
Altmetrics
Abstract
Many distributed systems, such as distributed ledger technologies, IoT technologies, and distributed databases must be able to handle participants that show unpredictable behavior. Unpredictable behavior can be a simple crash of a computer or a more severe situation where a participant is trying to maliciously manipulate the system. One fundamental problem in distributed systems is to agree on a common state by communicating over a peer-to-peer network. When malicious parties are present in the system, this problem is called Byzantine agreement. In order to solve Byzantine agreement, modern systems often assume that the communication channels are private and that therefore the messages communicated over the channels can not be seen by a malicious party.
This book revisits the original definition of Byzantine agreement where the communication channels are public. Protocols that solve Byzantine agreement over public channels are more powerful since malicious parties can access all messages that are communicated over all channels. This book transfers ideas of the original protocols to cope with real-life problems. In particular, it investigates whether it is possible for different height sensors to agree on a common height, whether candidates of an election can be ranked fairly even if malicious participants could overhear the ballots of others, under which assumptions Blockchain protocols can solve Byzantine agreement, and to what extent Reinforcement Learning can help us simulate and understand Byzantine behavior. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000451222Publication status
publishedExternal links
Search print copy at ETH Library
Publisher
ETH ZurichSubject
Distributed Computing; Byzantine agreement; Consensus; Asynchronous communication; Synchronous communication; Shared Memory; Message Passing; Multi-valued agreement; Multi-dimensional agreement; Validity properties; Preferential Voting; Append memory model; Kemeny Median; chain; DAG; Byzantine Reinforcement Learning; DRL; MULTI-AGENT SYSTEMS (ARTIFICIAL INTELLIGENCE); self-playOrganisational unit
03604 - Wattenhofer, Roger / Wattenhofer, Roger
More
Show all metadata
ETH Bibliography
yes
Altmetrics