Ensure presence and correctness of validity checks about ferveo HOT 16 CLOSED

• Ciphertext Validity Checking - must be done before creating a decryption share
• #38

tpke/src/iphertext.rs:

pub fn check_ciphertext_validity<E: PairingEngine>(
    c: &Ciphertext<E>,
    aad: &[u8],
) -> bool {
    let g_inv = E::G1Prepared::from(-E::G1Affine::prime_subgroup_generator());
    let hash_g2 = E::G2Prepared::from(construct_tag_hash::<E>(
        c.commitment,
        &c.ciphertext[..],
        aad,
    ));

    E::product_of_pairings(&[
        (E::G1Prepared::from(c.commitment), hash_g2),
        (g_inv, E::G2Prepared::from(c.auth_tag)),
    ]) == E::Fqk::one()
}

which is used by pub fn checked_decrypt and fn decrypt_with_shared_secret.

Also, in file tpke/src/api we have:

    pub fn to_decryption_share(&self) -> DecryptionShare {
        // TODO: Add verification steps

Verification is missing from this function.

from ferveo.

Yes, these are from the original Ferveo source code, and I'd like to figure out what they do, how to use them, whether we need them etc.

from ferveo.

• Verifying Decryption Shares - must be done before combining decryption shares

from ferveo.

• We don't use weights, but the indexing is still important

from ferveo.

• Are we going to be aggregating decryption shares?

from ferveo.

• Are we going to be aggregating decryption shares?

No, my understanding is that this makes sense for Ferveo original implementation, where a node (with a single private blinding key ek_i) may have many private shares, and the corresponding decryption shares can be aggregated.

from ferveo.

yeah, i couldn't think of any scenario where it would make sense for our use case

from ferveo.

I found some other checks that I would like to document and verify.

Moved to issue description

from ferveo.

oh nice, these are from the code and we need to verify the maths?

from ferveo.

We don't use weights, but the indexing is still important

Marking as complete since ferveo validators have a share_index attributed to them.

Are we going to be aggregating decryption shares?

Marking as solved.

I've aggregated the remaining checks into the issue description.

from ferveo.

Not sure how to compute 4.4.4 for a simple tDec variant. In this variant, $D_i$ is of type Fqk (product of pairing), but we need to be G1Prepared for $e(D_i, B_i)$ to work.

from ferveo.

verify_blinding doesn't work i.e. the original code seems to be implemented incorrectly. I can't find a reference to this operation in the docs or in the whitepaper.

@cygnusv, is this something we would like to explore further, or should we drop verify_blinding?

from ferveo.

If blinding_key, $B_i = [b] dk_i$, is the same as PublicKey::encryption_key of DKG validator, then verify_blinding in tpke is effectively the same check that verify_full in ferveo does:

• $e(G_1, \hat{Y}{i,\omega_j}) = e(A_{i,\omega_j}, ek_i)$ verify_full
• $e(g, Y_i) = e(A_i, [b] H)$, verify_blinding

If $B_i = [b] dk_i = b [H]$ doesn't hold, then I don't know what verify_blinding is supposed to check.

from ferveo.

Not sure how to compute 4.4.4 for a simple tDec variant. In this variant, Di is of type Fqk (product of pairing), but we need to be G1Prepared for e(Di,Bi) to work.

The naive rewrite doesn't make sense here anyway, since in simple tDec decryption share is defined as $C_i=e(U,Z_i)$, and in fast tDec as $D_i=[b−1]U$.

Marking this as a candidate for a research item in #42

from ferveo.

Found this earlier rewrite of verify_blinding.

from ferveo.

Closing this as the work will be continued in separate issues mentioned in OP.

from ferveo.