Cyber Security Seminar by Dr. Jérémie Decouchant - Accelerating blockchain consensus with trusted components
27 September 2022 12:00 till 12:45 - Location: Meeting Room B - Echo Building | Add to my calendar
Meeting Details
Location: Meetingroom B - Room-29.02.335 - ECHO building 2nd floor
Please note: 10 people can attend in person, further please join through zoom:
Meeting link (zoom)
Meeting ID: 928 2356 9831
Passcode: 272331
Abstract:
Consensus is a crucial building block of many distributed systems including state machine replication and blockchains. In particular, Byzantine Fault Tolerant (BFT) consensus protocols allow systems to withstand arbitrary failures. However, being able to tolerate arbitrary (Byzantine) component failures, induces a non-negligible system overhead, namely in the required number of nodes that should be present in the system as well as the communication complexity. For example, traditional BFT protocols, such as PBFT, typically require 3? +1 nodes to tolerate up to ? Byzantine faulty nodes.
Hybrid versions of traditional BFT protocols reduce the number of nodes required to ensure system safety as well as to decrease system latency. These protocols rely on secure elements that execute key functionalities in trusted execution environments such as inside Intel SGX enclaves. In this talk, we will present our recent works on hybrid protocols. These recent works modify consensus protocols that have been recently described and increase their resilience and performance. We will first discuss Damysus [EuroSys'22], which is a streamlined protocol that builds on HotStuff. We will then discuss a leaderless BFT protocol that leverages trusted components and that is ongoing work.
Short bio:
Jérémie Decouchant is an assistant professor at TU Delft in the Distributed Systems group and the Blockchain Lab since March 2021. He defended his PhD in 2016 from Grenoble University, France. Before moving to the Netherlands, he spent 5 years as a research associate (postdoc) and then as a research scientist at the Interdisciplinary Center for Security and Trust (SnT) in Luxembourg.
During his thesis, he focused on accountability and privacy in peer-to-peer content sharing systems. Since then, his research extended to neighboring areas and aims at defining resilient distributed algorithms, in particular for blockchain, federated learning and health data. At TU Delft, Jérémie is responsible for the Distributed Algorithms, and Operating Systems courses.