serai/crypto/dkg/src/encryption.rs

447 lines
14 KiB
Rust
Raw Normal View History

DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
use core::fmt::Debug;
use std::{
ops::Deref,
io::{self, Read, Write},
collections::HashMap,
};
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
use thiserror::Error;
use zeroize::{Zeroize, Zeroizing};
use rand_core::{RngCore, CryptoRng};
use chacha20::{
cipher::{crypto_common::KeyIvInit, StreamCipher},
Key as Cc20Key, Nonce as Cc20Iv, ChaCha20,
};
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
use transcript::{Transcript, RecommendedTranscript};
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
#[cfg(test)]
use group::ff::Field;
use group::GroupEncoding;
use ciphersuite::Ciphersuite;
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
use multiexp::BatchVerifier;
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
use schnorr::SchnorrSignature;
use dleq::DLEqProof;
use crate::ThresholdParams;
pub trait ReadWrite: Sized {
fn read<R: Read>(reader: &mut R, params: ThresholdParams) -> io::Result<Self>;
fn write<W: Write>(&self, writer: &mut W) -> io::Result<()>;
fn serialize(&self) -> Vec<u8> {
let mut buf = vec![];
self.write(&mut buf).unwrap();
buf
}
}
pub trait Message: Clone + PartialEq + Eq + Debug + Zeroize + ReadWrite {}
impl<M: Clone + PartialEq + Eq + Debug + Zeroize + ReadWrite> Message for M {}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
/// Wraps a message with a key to use for encryption in the future.
#[derive(Clone, PartialEq, Eq, Debug, Zeroize)]
pub struct EncryptionKeyMessage<C: Ciphersuite, M: Message> {
msg: M,
enc_key: C::G,
}
// Doesn't impl ReadWrite so that doesn't need to be imported
impl<C: Ciphersuite, M: Message> EncryptionKeyMessage<C, M> {
pub fn read<R: Read>(reader: &mut R, params: ThresholdParams) -> io::Result<Self> {
Ok(Self { msg: M::read(reader, params)?, enc_key: C::read_G(reader)? })
}
pub fn write<W: Write>(&self, writer: &mut W) -> io::Result<()> {
self.msg.write(writer)?;
writer.write_all(self.enc_key.to_bytes().as_ref())
}
pub fn serialize(&self) -> Vec<u8> {
let mut buf = vec![];
self.write(&mut buf).unwrap();
buf
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
// Used by tests
pub(crate) fn enc_key(&self) -> C::G {
self.enc_key
}
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
pub trait Encryptable: Clone + AsRef<[u8]> + AsMut<[u8]> + Zeroize + ReadWrite {}
impl<E: Clone + AsRef<[u8]> + AsMut<[u8]> + Zeroize + ReadWrite> Encryptable for E {}
/// An encrypted message, with a per-message encryption key enabling revealing specific messages
/// without side effects.
#[derive(Clone, Zeroize)]
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
pub struct EncryptedMessage<C: Ciphersuite, E: Encryptable> {
key: C::G,
// Also include a proof-of-possession for the key.
// If this proof-of-possession wasn't here, Eve could observe Alice encrypt to Bob with key X,
// then send Bob a message also claiming to use X.
// While Eve's message would fail to meaningfully decrypt, Bob would then use this to create a
// blame argument against Eve. When they do, they'd reveal bX, revealing Alice's message to Bob.
// This is a massive side effect which could break some protocols, in the worst case.
// While Eve can still reuse their own keys, causing Bob to leak all messages by revealing for
// any single one, that's effectively Eve revealing themselves, and not considered relevant.
pop: SchnorrSignature<C>,
msg: Zeroizing<E>,
}
fn ecdh<C: Ciphersuite>(private: &Zeroizing<C::F>, public: C::G) -> Zeroizing<C::G> {
Zeroizing::new(public * private.deref())
}
// Each ecdh must be distinct. Reuse of an ecdh for multiple ciphers will cause the messages to be
// leaked.
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
fn cipher<C: Ciphersuite>(dst: &'static [u8], ecdh: &Zeroizing<C::G>) -> ChaCha20 {
// Ideally, we'd box this transcript with ZAlloc, yet that's only possible on nightly
// TODO: https://github.com/serai-dex/serai/issues/151
let mut transcript = RecommendedTranscript::new(b"DKG Encryption v0.2");
transcript.domain_separate(dst);
let mut ecdh = ecdh.to_bytes();
transcript.append_message(b"shared_key", ecdh.as_ref());
ecdh.as_mut().zeroize();
let zeroize = |buf: &mut [u8]| buf.zeroize();
let mut key = Cc20Key::default();
let mut challenge = transcript.challenge(b"key");
key.copy_from_slice(&challenge[.. 32]);
zeroize(challenge.as_mut());
// Since the key is single-use, it doesn't matter what we use for the IV
// The isssue is key + IV reuse. If we never reuse the key, we can't have the opportunity to
// reuse a nonce
// Use a static IV in acknowledgement of this
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
let mut iv = Cc20Iv::default();
// The \0 is to satisfy the length requirement (12), not to be null terminated
iv.copy_from_slice(b"DKG IV v0.2\0");
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
// ChaCha20 has the same commentary as the transcript regarding ZAlloc
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
// TODO: https://github.com/serai-dex/serai/issues/151
let res = ChaCha20::new(&key, &iv);
zeroize(key.as_mut());
res
}
fn encrypt<R: RngCore + CryptoRng, C: Ciphersuite, E: Encryptable>(
rng: &mut R,
dst: &'static [u8],
from: u16,
to: C::G,
mut msg: Zeroizing<E>,
) -> EncryptedMessage<C, E> {
/*
The following code could be used to replace the requirement on an RNG here.
It's just currently not an issue to require taking in an RNG here.
let last = self.last_enc_key.to_bytes();
self.last_enc_key = C::hash_to_F(b"encryption_base", last.as_ref());
let key = C::hash_to_F(b"encryption_key", last.as_ref());
last.as_mut().zeroize();
*/
// Generate a new key for this message, satisfying cipher's requirement of distinct keys per
// message, and enabling revealing this message without revealing any others
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
let key = Zeroizing::new(C::random_nonzero_F(rng));
cipher::<C>(dst, &ecdh::<C>(&key, to)).apply_keystream(msg.as_mut().as_mut());
let pub_key = C::generator() * key.deref();
let nonce = Zeroizing::new(C::random_nonzero_F(rng));
let pub_nonce = C::generator() * nonce.deref();
EncryptedMessage {
key: pub_key,
pop: SchnorrSignature::sign(
&key,
nonce,
pop_challenge::<C>(pub_nonce, pub_key, from, msg.deref().as_ref()),
),
msg,
}
}
impl<C: Ciphersuite, E: Encryptable> EncryptedMessage<C, E> {
pub fn read<R: Read>(reader: &mut R, params: ThresholdParams) -> io::Result<Self> {
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
Ok(Self {
key: C::read_G(reader)?,
pop: SchnorrSignature::<C>::read(reader)?,
msg: Zeroizing::new(E::read(reader, params)?),
})
}
pub fn write<W: Write>(&self, writer: &mut W) -> io::Result<()> {
writer.write_all(self.key.to_bytes().as_ref())?;
self.pop.write(writer)?;
self.msg.write(writer)
}
pub fn serialize(&self) -> Vec<u8> {
let mut buf = vec![];
self.write(&mut buf).unwrap();
buf
}
#[cfg(test)]
pub(crate) fn invalidate_pop(&mut self) {
self.pop.s += C::F::one();
}
#[cfg(test)]
pub(crate) fn invalidate_msg<R: RngCore + CryptoRng>(&mut self, rng: &mut R, from: u16) {
// Invalidate the message by specifying a new key/Schnorr PoP
// This will cause all initial checks to pass, yet a decrypt to gibberish
let key = Zeroizing::new(C::random_nonzero_F(rng));
let pub_key = C::generator() * key.deref();
let nonce = Zeroizing::new(C::random_nonzero_F(rng));
let pub_nonce = C::generator() * nonce.deref();
self.key = pub_key;
self.pop = SchnorrSignature::sign(
&key,
nonce,
pop_challenge::<C>(pub_nonce, pub_key, from, self.msg.deref().as_ref()),
);
}
// Assumes the encrypted message is a secret share.
#[cfg(test)]
pub(crate) fn invalidate_share_serialization<R: RngCore + CryptoRng>(
&mut self,
rng: &mut R,
dst: &'static [u8],
from: u16,
to: C::G,
) {
use group::ff::PrimeField;
let mut repr = <C::F as PrimeField>::Repr::default();
for b in repr.as_mut().iter_mut() {
*b = 255;
}
// Tries to guarantee the above assumption.
assert_eq!(repr.as_ref().len(), self.msg.as_ref().len());
// Checks that this isn't over a field where this is somehow valid
assert!(!bool::from(C::F::from_repr(repr).is_some()));
self.msg.as_mut().as_mut().copy_from_slice(repr.as_ref());
*self = encrypt(rng, dst, from, to, self.msg.clone());
}
// Assumes the encrypted message is a secret share.
#[cfg(test)]
pub(crate) fn invalidate_share_value<R: RngCore + CryptoRng>(
&mut self,
rng: &mut R,
dst: &'static [u8],
from: u16,
to: C::G,
) {
use group::ff::PrimeField;
// Assumes the share isn't randomly 1
let repr = C::F::one().to_repr();
self.msg.as_mut().as_mut().copy_from_slice(repr.as_ref());
*self = encrypt(rng, dst, from, to, self.msg.clone());
}
}
/// A proof that the provided point is the legitimately derived shared key for some message.
#[derive(Clone, PartialEq, Eq, Debug, Zeroize)]
pub struct EncryptionKeyProof<C: Ciphersuite> {
key: Zeroizing<C::G>,
dleq: DLEqProof<C::G>,
}
impl<C: Ciphersuite> EncryptionKeyProof<C> {
pub fn read<R: Read>(reader: &mut R) -> io::Result<Self> {
Ok(Self { key: Zeroizing::new(C::read_G(reader)?), dleq: DLEqProof::read(reader)? })
}
pub fn write<W: Write>(&self, writer: &mut W) -> io::Result<()> {
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
writer.write_all(self.key.to_bytes().as_ref())?;
self.dleq.write(writer)
}
pub fn serialize(&self) -> Vec<u8> {
let mut buf = vec![];
self.write(&mut buf).unwrap();
buf
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
#[cfg(test)]
pub(crate) fn invalidate_key(&mut self) {
*self.key += C::generator();
}
#[cfg(test)]
pub(crate) fn invalidate_dleq(&mut self) {
let mut buf = vec![];
self.dleq.write(&mut buf).unwrap();
// Adds one to c since this is serialized c, s
// Adding one to c will leave a validly serialized c
// Adding one to s may leave an invalidly serialized s
buf[0] = buf[0].wrapping_add(1);
self.dleq = DLEqProof::read::<&[u8]>(&mut buf.as_ref()).unwrap();
}
}
// This doesn't need to take the msg. It just doesn't hurt as an extra layer.
// This still doesn't mean the DKG offers an authenticated channel. The per-message keys have no
// root of trust other than their existence in the assumed-to-exist external authenticated channel.
fn pop_challenge<C: Ciphersuite>(nonce: C::G, key: C::G, sender: u16, msg: &[u8]) -> C::F {
let mut transcript = RecommendedTranscript::new(b"DKG Encryption Key Proof of Possession v0.2");
transcript.append_message(b"nonce", nonce.to_bytes());
transcript.append_message(b"key", key.to_bytes());
// This is sufficient to prevent the attack this is meant to stop
transcript.append_message(b"sender", sender.to_le_bytes());
// This, as written above, doesn't hurt
transcript.append_message(b"message", msg);
// While this is a PoK and a PoP, it's called a PoP here since the important part is its owner
// Elsewhere, where we use the term PoK, the important part is that it isn't some inverse, with
// an unknown to anyone discrete log, breaking the system
C::hash_to_F(b"DKG-encryption-proof_of_possession", &transcript.challenge(b"schnorr"))
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
fn encryption_key_transcript() -> RecommendedTranscript {
RecommendedTranscript::new(b"DKG Encryption Key Correctness Proof v0.2")
}
#[derive(Clone, Copy, PartialEq, Eq, Debug, Error)]
pub(crate) enum DecryptionError {
#[error("accused provided an invalid signature")]
InvalidSignature,
#[error("accuser provided an invalid decryption key")]
InvalidProof,
}
// A simple box for managing encryption.
#[derive(Clone)]
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
pub(crate) struct Encryption<C: Ciphersuite> {
dst: &'static [u8],
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
i: u16,
enc_key: Zeroizing<C::F>,
enc_pub_key: C::G,
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
enc_keys: HashMap<u16, C::G>,
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
impl<C: Ciphersuite> Zeroize for Encryption<C> {
fn zeroize(&mut self) {
self.enc_key.zeroize();
self.enc_pub_key.zeroize();
for (_, mut value) in self.enc_keys.drain() {
value.zeroize();
}
}
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
impl<C: Ciphersuite> Encryption<C> {
pub(crate) fn new<R: RngCore + CryptoRng>(dst: &'static [u8], i: u16, rng: &mut R) -> Self {
let enc_key = Zeroizing::new(C::random_nonzero_F(rng));
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
Self {
dst,
i,
enc_pub_key: C::generator() * enc_key.deref(),
enc_key,
enc_keys: HashMap::new(),
}
}
pub(crate) fn registration<M: Message>(&self, msg: M) -> EncryptionKeyMessage<C, M> {
EncryptionKeyMessage { msg, enc_key: self.enc_pub_key }
}
pub(crate) fn register<M: Message>(
&mut self,
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
participant: u16,
msg: EncryptionKeyMessage<C, M>,
) -> M {
if self.enc_keys.contains_key(&participant) {
panic!("Re-registering encryption key for a participant");
}
self.enc_keys.insert(participant, msg.enc_key);
msg.msg
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
pub(crate) fn encrypt<R: RngCore + CryptoRng, E: Encryptable>(
&self,
rng: &mut R,
participant: u16,
msg: Zeroizing<E>,
) -> EncryptedMessage<C, E> {
encrypt(rng, self.dst, self.i, self.enc_keys[&participant], msg)
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
pub(crate) fn decrypt<R: RngCore + CryptoRng, I: Copy + Zeroize, E: Encryptable>(
&self,
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
rng: &mut R,
batch: &mut BatchVerifier<I, C::G>,
// Uses a distinct batch ID so if this batch verifier is reused, we know its the PoP aspect
// which failed, and therefore to use None for the blame
batch_id: I,
from: u16,
mut msg: EncryptedMessage<C, E>,
) -> (Zeroizing<E>, EncryptionKeyProof<C>) {
msg.pop.batch_verify(
rng,
batch,
batch_id,
msg.key,
pop_challenge::<C>(msg.pop.R, msg.key, from, msg.msg.deref().as_ref()),
);
let key = ecdh::<C>(&self.enc_key, msg.key);
cipher::<C>(self.dst, &key).apply_keystream(msg.msg.as_mut().as_mut());
(
msg.msg,
EncryptionKeyProof {
key,
dleq: DLEqProof::prove(
rng,
&mut encryption_key_transcript(),
&[C::generator(), msg.key],
&self.enc_key,
),
},
)
}
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
// Given a message, and the intended decryptor, and a proof for its key, decrypt the message.
// Returns None if the key was wrong.
pub(crate) fn decrypt_with_proof<E: Encryptable>(
&self,
DKG Blame (#196) * Standardize the DLEq serialization function naming They mismatched from the rest of the project. This commit is technically incomplete as it doesn't update the dkg crate. * Rewrite DKG encryption to enable per-message decryption without side effects This isn't technically true as I already know a break in this which I'll correct for shortly. Does update documentation to explain the new scheme. Required for blame. * Add a verifiable system for blame during the FROST DKG Previously, if sent an invalid key share, the participant would realize that and could accuse the sender. Without further evidence, either the accuser or the accused could be guilty. Now, the accuser has a proof the accused is in the wrong. Reworks KeyMachine to return BlameMachine. This explicitly acknowledges how locally complete keys still need group acknowledgement before the protocol can be complete and provides a way for others to verify blame, even after a locally successful run. If any blame is cast, the protocol is no longer considered complete-able (instead aborting). Further accusations of blame can still be handled however. Updates documentation on network behavior. Also starts to remove "OnDrop". We now use Zeroizing for anything which should be zeroized on drop. This is a lot more piece-meal and reduces clones. * Tweak Zeroizing and Debug impls Expands Zeroizing to be more comprehensive. Also updates Zeroizing<CachedPreprocess([u8; 32])> to CachedPreprocess(Zeroizing<[u8; 32]>) so zeroizing is the first thing done and last step before exposing the copy-able [u8; 32]. Removes private keys from Debug. * Fix a bug where adversaries could claim to be using another user's encryption keys to learn their messages Mentioned a few commits ago, now fixed. This wouldn't have affected Serai, which aborts on failure, nor any DKG currently supported. It's just about ensuring the DKG encryption is robust and proper. * Finish moving dleq from ser/deser to write/read * Add tests for dkg blame * Add a FROST test for invalid signature shares * Batch verify encrypted messages' ephemeral keys' PoP
2023-01-01 06:54:18 +00:00
from: u16,
decryptor: u16,
mut msg: EncryptedMessage<C, E>,
// There's no encryption key proof if the accusation is of an invalid signature
proof: Option<EncryptionKeyProof<C>>,
) -> Result<Zeroizing<E>, DecryptionError> {
if !msg
.pop
.verify(msg.key, pop_challenge::<C>(msg.pop.R, msg.key, from, msg.msg.deref().as_ref()))
{
Err(DecryptionError::InvalidSignature)?;
}
if let Some(proof) = proof {
// Verify this is the decryption key for this message
proof
.dleq
.verify(
&mut encryption_key_transcript(),
&[C::generator(), msg.key],
&[self.enc_keys[&decryptor], *proof.key],
)
.map_err(|_| DecryptionError::InvalidProof)?;
cipher::<C>(self.dst, &proof.key).apply_keystream(msg.msg.as_mut().as_mut());
Ok(msg.msg)
} else {
Err(DecryptionError::InvalidProof)
}
}
}