Statistical Bug Isolation for Consensus Systems
More Info
expand_more
expand_more
Abstract
The testing of consensus systems has received growing attention and recent testing tools generate many faulty executions. However, there is a lack of methods that automatically analyze these outputs to identify the root causes of the bugs they found.
This paper presents Isolation, a statistical bug isolation algorithm that uses message-based predicates to discriminate between different faults. We applied our method to executions of the XRP Ledger blockchain and evaluated its performance. Our comparison shows that Isolation correctly separates bugs by their root cause and consistently outperforms the state-of-the-art with higher F-scores.
Files
Thesis.pdf
(.pdf | 0.399 Mb)