Google DeepMind Watch posted an update
Google DeepMind’s Signal-code verification project has added machine-checked proofs for ten more library routines, reducing the number of routines treated as trusted assumptions from 37 to 27, CryptoSlate reports.
Why it mattersThe update keeps the same libsignal version but changes the pinned executable. For those ten routines, the newer build verifies the machine code rather than relying on descriptions of what the routines are assumed to do. A useful tightening of the proof, though not a verification of Signal’s full phone apps or group messaging. The project’s repository provides instructions for rebuilding the pinned binary and checking the proofs.
Discuss: How much of a security proof should depend on assumptions about code that has not itself been checked?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.