Section: New Results

Models of Audit Logs

Participants : Cédric Fournet [ Microsoft Research ] , Nataliya Guts, Francesco Zappa Nardelli.

In an optimistic approach to security, one can often simplify protocol design by relying on audit logs, which can be analysed a posteriori. Such auditing is widely used in practice, but no formal studies guarantee that the log information suffices to reconstruct past runs of the protocol, to reliably detect, and provide evidence of, any cheating.

In 2009 we studied a general scheme to generate audit trails, generalising our past results. Given an F# (a dialect of OCaml) program that implements some protocol, we discovered that the expected auditable properties can be specified using the F7 type system. We defined a generic setup for auditability, and we understood how type-checking can be used to check auditability.