A Novel Evaluation Platform for Measuring Endurance of Byzantine Fault-Tolerant Systems under Injected Stressors (NEMESIS)
Third party funded individual grant
Acronym: NEMESIS
Start date : 01.01.2027
End date : 31.12.2029
Project details
Short description
Byzantine fault-tolerant (BFT) consensus protocols nowadays form the backbone of Distributed Ledger Technology (DLT) systems. A new generation of BFT protocols organises blocks in a directed acyclic graph instead of a linear chain, thereby achieving substantially higher throughput and better scalability. These protocols already secure considerable economic value in production DLTs. However, their attack surface and their robustness under realistic conditions remain insufficiently explored.
Existing evaluation tools are mostly island solutions: they do not scale to realistic system sizes, require re-implementing the protocols, fail to realistically reproduce wide-area networks (WANs), or do not yield reproducible results. NEMESIS closes this gap by (1) developing a scalable, realistic, and reproducible evaluation platform for BFT protocols, (2) designing search techniques for coordinated Byzantine attacks, and (3) providing methods for detecting liveness issues.
The project investigates how to design a modular platform that combines the realism of emulation (execution of unmodified protocol binaries) with the scalability and determinism of simulation methods, thus pursuing a hybrid emulation/simulation approach. The platform executes protocol implementations written in different programming languages, with as few modifications as possible, inside a deterministic simulation environment. It recreates WAN conditions including latency and bandwidth constraints, scales to 100 and more nodes, and allows protocol runs to be replayed exactly. Building on this, NEMESIS models a strategic adversary that orchestrates all Byzantine nodes and exploits knowledge of protocol roles and states to explore the vast space of possible attacks in a targeted manner. Besides classical attacks on consistency such as equivocation, NEMESIS also covers volume-based attacks that target resource exhaustion or flawed garbage collection.
Finally, NEMESIS combines analytical and empirical techniques to detect both hard liveness violations and states in which a protocol formally guarantees progress but becomes practically unusable, for instance due to stalling block production or severely increased latency. All results, in particular the platform itself, will be released as open source. Overall, NEMESIS aims to bridge the gap between formal correctness guarantees and the operational usability of BFT systems by developing techniques that make practically relevant liveness issues visible and measurable.
Scientific Abstract
Byzantine fault-tolerant (BFT) consensus protocols nowadays form the backbone of Distributed Ledger Technology (DLT) systems. A new generation of BFT protocols organises blocks in a directed acyclic graph instead of a linear chain, thereby achieving substantially higher throughput and better scalability. These protocols already secure considerable economic value in production DLTs. However, their attack surface and their robustness under realistic conditions remain insufficiently explored.
Existing evaluation tools are mostly island solutions: they do not scale to realistic system sizes, require re-implementing the protocols, fail to realistically reproduce wide-area networks (WANs), or do not yield reproducible results. NEMESIS closes this gap by (1) developing a scalable, realistic, and reproducible evaluation platform for BFT protocols, (2) designing search techniques for coordinated Byzantine attacks, and (3) providing methods for detecting liveness issues.
The project investigates how to design a modular platform that combines the realism of emulation (execution of unmodified protocol binaries) with the scalability and determinism of simulation methods, thus pursuing a hybrid emulation/simulation approach. The platform executes protocol implementations written in different programming languages, with as few modifications as possible, inside a deterministic simulation environment. It recreates WAN conditions including latency and bandwidth constraints, scales to 100 and more nodes, and allows protocol runs to be replayed exactly. Building on this, NEMESIS models a strategic adversary that orchestrates all Byzantine nodes and exploits knowledge of protocol roles and states to explore the vast space of possible attacks in a targeted manner. Besides classical attacks on consistency such as equivocation, NEMESIS also covers volume-based attacks that target resource exhaustion or flawed garbage collection.
Finally, NEMESIS combines analytical and empirical techniques to detect both hard liveness violations and states in which a protocol formally guarantees progress but becomes practically unusable, for instance due to stalling block production or severely increased latency. All results, in particular the platform itself, will be released as open source. Overall, NEMESIS aims to bridge the gap between formal correctness guarantees and the operational usability of BFT systems by developing techniques that make practically relevant liveness issues visible and measurable.
