Sites Inria

English version

Séminaire des équipes de recherche

Séminaire Gallium : EverParse: verified efficient parsing for transport data formats

© INRIA Sophie Auvin - G comme Grille

Parsers are a security-critical part of networking software, where bugs can lead to such catastrophes as the 2017 Cloudflare data leaks ("Cloudbleed".) 

We present EverParse, our open-source library and tools for automatic generation of verified efficient parsers for binary transport data formats.

  • Date : 3/10/2019
  • Lieu : Salle Jacques-Louis Lions 1, bâtiment C - 10h30
  • Intervenant(s) : Tahina Ramananandro (Microsoft Research, Redmond)

EverParse takes a data format description and produces an F* implementation of parsers and serializers for it. EverParse produces low-level code in the Low* subset of F* that is then compiled to C. The produced code is verified for memory safety, functional correctness (parsers and serializers are inverse of each other) and such security properties as non-malleability (unique binary representation for authentication purposes.)

The produced code is memory-efficient, minimizing the need for memory allocations and copies. EverParse is part of Project Everest, a collaborative (MSR-CMU-INRIA-U.Edinburgh) research effort to deploy verified implementations of the TLS protocol and derivatives.

Mots-clés : Data Recherche Inria de Paris Gallium Séminaire Transport Data Parsing

Haut de page

Suivez Inria