FORMAL VERIFICATIONComing soon

Formal verification
of the handshake.

Not done yet. What it must prove, who will do it and what will be published here.

STATUS

Where it stands.

Status
The modelNot built yet.
ToolTamarin or ProVerif, chosen by the specialist.
WhoAn outside specialist. Not engaged yet.
WhenBefore the handshake is public.
Published hereThe model, its assumptions and its results, once they exist.

WHAT IT MUST PROVE

Four properties.

Mutual authentication

When a session opens, each side really talked to the agent its credential names, and both agree on who that is.

Secrecy of the session binding

No one else can make requests that count as part of the session.

Freshness

No old message can be replayed into a new exchange.

Downgrade resistance

No one can push the two sides to a lower version than both speak.

ITS ASSUMPTIONS

What the model will assume.

Binding in the application

Version 0.1 binds the exchange with nonces in the application, not to the TLS connection. The model has to say so.

Relays

A party that ends TLS for one side can pass messages on unchanged. The model has to show that this gains it nothing it can use.

Keys

Each agent’s key is its own. A stolen key is handled by expiry, revocation and freezing, not by the handshake.

WHAT EXISTS TODAY

Tests, not a proof.

The platform’s tests run a test vector for every result code, and the code is checked against them byte for byte. Tests show the code does what the specification says. They do not show the specification is sound: that is the model’s job.

The handshake