NOTE: I am in the process of updating this implementation to match a new spec. If you found this after I sent you an RFC please don't go spelunking because this implementation has much weaker privacy properties than what I'm attempting to accomplish with the RFC.
Blizzard is a protocol for setting up authenticated and encrypted channels. It is designed to be used in peer-to-peer networks where peers are not necessarily online at the same time i.e. it is a delay-tolerant protocol. Unlike most other DTN protocols, it's designed to be used on a microcontroller. This crate is
no_std-compatible, but does require an allocator.
Blizzard is built on top of a
no_std-compatible fork of snow, a noise protocol implementation. Blizzard uses the noise protocol framework to handle encryption and authentication, meaning there are zero (or perhaps one) bespoke usages of cryptographic primitives in this crate. The closest thing is the use of a key deriviation function to turn a noise public key into a 64-bit fingerprint. This fingerprint is used to address messages to a specific peer.
Blizzard does not handle any transport layer concerns. It is up to the user to send and receive messages over a transport. Blizzard messages can be sent over any transport, including unreliable transports like UDP, but it is currently up to the user to handle message ordering and retransmission. Noise does have facilities for handling these concerns, but Blizzard does not currently use them. This may change in the future.
Blizzard is not tolerant of sybil attacks, black-holing, or other malicious denial-of-service type behavior. Network operation is predicated on the assumption that peers are honest and will not attempt to disrupt the network, however bad actors should not be capable of compromising the security or anonymity guarantees of the protocol.
Forwarding logic has not yet been defined, so at the moment Blizzard is is only suitable for direct peer-to-peer connections.
To the best of my knowledge, blizzard offers initiator anonymity, forward secrecy (each session uses ephemeral keys), authenticated and encrypted channels via the noise protocol framework, and some amount of plausible deniability of network topology. Blizzard does not offer information-theoretic anonymity guarantees for the recipient of a message, meaning packets are addressed to a specific peer who may be forced to reveal an association with the destination address in order to trigger delivery. Announces are built in a way that intentionally obfuscates network topology, meaning it may be difficult for an attacker to determine whether a node is the recipient of a message or merely a relay. This is not a guarantee of anonymity.
The main feature of Blizzard, though, is that it's a DTN protocol built for humans to understand. The entire protocol as of this writing is less than 1000 lines of code, including tests and documentation. The implementation itself is closer to 500 lines, and doesn't do anything scary. Building on top of
snow is what makes this possible.
Blizzard is not ready for general use. The API should not be considered stable. It has not been simulated. It has not been audited.
Blizzard is dual-licensed under Apache 2.0 and MIT.