SigmaProofs is an ultimate swiss knife for implementing cryptogrpahic protocols that uses group based zero knowledge proofs often having commitment, challenge and reply structure for the proofs. This package includes a Verificatum compatable parser which is easuy to use for writting challenges and doing serialization/deserialization of a whole simulator which can often be a rather daunting task. Easy to use ElGamal implementation is also made necessary for reencryption mixnets as further used by ShuffleProofs.jl.
The provided proofs are made verifier agnostic making it easy to explore the specification space for writing challenges for the proofs and implement different generator basis algorithms. Every proof is treated as simulator that consists of three parts:
Proposition: the statement that is being proved with a followiong proof typeProof: a proof that supports the statementVerifier: a verifier that participates in proof generation and can attest validity of the proof
The Verifier is often made noninteractive via FiatShamir transformation nevertheless if one wishes interacive proofs are also supported. This is possible to do via the same codebas as Julia does not colour functions with @async as many other programing languages do (although you would probably not need this).
The methods that are uniformly implemnted are prove(::Proposition, secrets..., ::Verifier)::Proof, challenge(::Verifier, ::Proposition, args...), verify(::Proposition, ::Proof, ::Verifier)::Bool and proof_type(::Type{<:Proposition})::Type{<:Proof}. To make the proofs convinient to use most of the proposition types expose construction methods that has semantics construct(args..., secrets..., ::Verifier)::Simulator like for instance exponentiate(g, sk, verifier). The simulator has fields proposition, proof, verifier and can be convininiently verified via verify(::Simulator)::Bool method. In addition there is generator_basis(::Verifier, ::Type{<:Group}, N)::Vector{G} that produces independent generator basis vector particulalry useful for generalized pedersen commitments.
The SigmaProofs implements a Verificatum inspired verifier that provides the same Verifier semantics, uses the same generator basis algorithm as in Verificatum specification. The challenges to the proofs with the verifier are constructed using binary tree parser as given in the specification implemnted in Parsers module. However, it is important to not that only ShuffleProofs is implemented to be compatable with the Verificatum psecification as other proof types are not considered there. This package can be looked as a one possible way for verifier extension to other proof types.
LogProofsThe basis of the zero knowledge, for proving statementsPK{(x): y <- g^x}which is known in the literature as SchnorrProof used for proving the knowledge of an exponent linking two public group elements(g, y). On it's basis via adding asuffixmessage to the challenge a cryptographic signature is made known as SchnorrSignature. The other type of the proof isPK{(x): A_i y_i <- g_i^x }which proves that a lsit of pairs(g_i, y_i)has the same exponent known as ChaumPedersenProof.
using CryptoGroups
using SigmaProofs.LogProofs: exponentiate, verify
using SigmaProofs.Verificatum
g = @ECGRoup{P_192}()
verifier = ProtocolSpec(; g)
sk = 42
pk = g^sk
g_vec = [g, g^2, g^5]
simulator = exponentiate(g_vec, sk, verifier)
verify(simulator) # true
simulator.proposition.y == [pk, pk^2, pk^5] # trueElGamal cypherrtext is encoded in ElGamalRow type which can be initialized with (a, b) either being elements of a group or an a tuple of elements representing generalized ElGamal cypheretxt that can extend the amount of information that can be encoded. When initialized with a single group element or a tupel it is presumed to be a message. The type suppports width method telling how wide the encoded cyphertext is.
ElGamalRow elements can conviniently be reencrypted with encryptor initialized with Enc type and then also decrypted with Dec. Convineitly the encryptor can also be used to initialize the cyphertexts.
using CryptoGroups
g = @ECGRoup{P_192}()
sk = 42
pk = g^sk
enc = Enc(pk, g)
dec = Dec(sk)
plaintexts = (g^2, g^5)
r = (8, 10)
ciphertext = enc(plaintexts, r)
ciphertext == ElGamalRow(g .^ r, pk .^r .* m) # true
dec(ciphertext) == plaintexts #
width(ciphertext) == 2
# Arithmetic operations `*`, `^` and `/` are also supported
dec(ciphertext * cyphertext) == plaintexts .* plaintexts
dec(ciphertext ^ 7) == plaintexts .^ 7
dec(ciphertext / enc(ciphertext, r)) == (one(g), one(g))The latter division can be used as plaintext equivalence test for the package. The arithmetic operations are convinient to make some zero knowledge proof implementations more coincise.
DecryptionProofs takes CahumPedersenProofs and uses them to construct ElGamalRow cypheretext vector and can produce coresponding zero knowledge proofs of correct decryption:
using CryptoGroups
using SigmaProofs.ElGamal: Enc
using SigmaProofs.DecryptionProofs: exponentiate, verify
g = @ECGroup{P_192}()
verifier = ProtocolSpec(; g)
sk = 123
pk = g^sk
encryptor = Enc(pk, g)
plaintexts = [g^4, g^2, g^3] .|> tuple
cyphertexts = encryptor(plaintexts, rand(2:order(g)-1, length(plaintexts)))
simulator = decrypt(g, cyphertexts, sk, verifier)
verify(simulator) # true
simultor.proposition.plaintexts == plaintextsRange proofs allow one tow prove to the verifier that an integer within a certain cyphertext or a commit is within a given range without revealing it in the first place. The most simple method is bitenc that can produce proof that given elgamal cyphertex is either (g^r, pk^r * 𝐺) OR (g^r, pk^r / 𝐺) without revealing r.
simulator = bitenc(g, pk, true, verifier)
@test verify(simulator)
the proposition ElGamalBitRange contains fields x and y that are elgamal encryptions. The proof is taken directly from a paper:
- [1]: Ronald Cramer, Rosario Gennaro, and Berry Schoenmakers. 1997. A secure and optimally efficient multi-authority election scheme. In Proceedings of the 16th annual international conference on Theory and application of cryptographic techniques (EUROCRYPT'97). Springer-Verlag, Berlin, Heidelberg, 103–118. https://link.springer.com/chapter/10.1007/3-540-69053-0_9
To commit arbitrary values n-arry decomposition is used. It is possible to reduce used group elements making the proof for commits only. For that we have a bitcommit(g::G, h::G, value::Bool, verifier::Verifier)::Simulator function which is analougous to bitenc except now h must be indeoendent from g and produces pedersen commitment in the result.
The bitcommit function is generalized with n-ary decomposition into rangecommit(range::Union{Int, UnitRange}, g::G, h::G, value::Int, ::Verifier). When the range is integer then the range is assumed to be 0 <= value < 2^n. Arbitray ranges can be useful for identity based signatures. The general idea is that client generates and identity provider signs your commitment C = h^β * g^v. The value can encode birth date or latitude/longtitude which can allow to prove to a service provider that your age is over 18 or that you live in a particular contry via issuing a requested range proof. Such setup prevents the identity provider to learn when or what you have accessed.
Another application of range commitments is that they can be used in a sealed bid auctions. An intersting variant is a Vickrey auction where the highest bidder wins but pasy the second-highest bid price. In such scheme the bidder signs a commitment and sends that to the dealer along with oppening. The dealer publishes the commitment on the buletin baord while keeps signature and oppening a secret. When all bids <- h^β * g^v are collected the dealer can anounce and prove the winner via running
simulator = vickrey_auction(bids, g, h, 𝛃, verifier)`
verify(simulator) # true
(; winner, value_win) = simulator.proposition
where the winner is the index in the list of bids and value_win is the winning value. The contract obligation then can be enforced by the dealer anouncing the signature while leaving other bidders' identitites and their bidded values secret. For full example see examples/auctions.jl.
The last piece of puzzle is range proof use in homomorphic tallying methods using in some evoting system designs. Here the vote is ElGamal encrypted and then homomoprhically talllied with only resulting tally being decrypted in trheshold decryption ceremony without revelaing individual votes. For this process to work, however, there is a need to ensure that voters had cast their votes honestly within allowed range. For simple yes/no referenda one would use bitenc wheras for multiple options one can make range encryption proofsexposed with rangeenc(range::Union{Int, UnitRange}, g::G, pk::G, value::Int, ::Verifier; 𝐺)::Simulator which uses rangecommit and also provides proof of correct ElGamal encryption where used randmization factors for a and b are the same.
The SigmaProofs.SecretSharing module provides an implementation of the Feldman Verifiable Secret Sharing (VSS) scheme, a cryptographic protocol that extends Shamir's Secret Sharing with a verification mechanism.
The Feldman VSS scheme allows a dealer to distribute shares of a secret among participants in a way that enables the participants to verify the consistency of their shares without revealing the secret. This is achieved by the dealer publishing commitments to the coefficients of the polynomial used to generate the shares.
The module includes the following key functionality:
-
Secret Sharing Setup: The
sharding_setup(g::G, nodes::Vector{G}, coeff::Vector{G}, ::Verifier)::Simulator{ShardingSetup}function generates the necessary components for the secret sharing scheme, including the dealer's public key, the participants' public keys, and the commitments to the polynomial coefficients. -
Share Verification: The
verifyfunction allows participants to verify the consistency of their shares by checking the published commitments. -
Secret Reconstruction: The
merge_exponentiations(::Simulator{ShardingSetup}, ::Vector{G}, ::Vector{Exponentiation{G}}, proof::Vector{ChaumPedersenProof{G}})::Simulator{Exponentiation}andmerge_decryptions(::Simulator{ShardingSetup}, ::Vector{<:ElGamalRow{G}}, ::Vector{Exponentiation{G}}, proof::Vector{ChaumPedersenProof{G}})::Simulator{Decryption}functions enable the reconstruction of the secret by combining a threshold number of shares using Lagrange interpolation. -
Utility Functions: The module provides helper functions for polynomial evaluation (
evaluate_poly) and Lagrange coefficient computation (lagrange_coef), which are essential for both the sharing and reconstruction processes.
The module also includes support for generating proofs and verifying the consistency of the scheme using the SigmaProofs framework. For more detailed use see test/secretsharing.jl file.
Commitment shuffle lays out a new shuffle protocol (I am not aware of it's previous uses) that uses generalized pedersen commitments. In contrast to ShuffleProofs it is interactive, however, interactivity is asynchronous and it provides information-theoretic security for a shuffle that is sometimes needed fto ensuring ever lasting privacy in case a new cryptograhic breaking algorithm surfaces or cryptographic relevant qauntuim computer ever gets built in the future. Another benefit of the shuffle is that it's outputs are unstructured and can contain anything allowing easy cryptographic protocol composition. The procol is:
-
Setup:
- A verifiable generator list is established
- A blinding generator 'h' is set
-
Dealer-member interaction:
- Dealer sends a verifiably random generator g_i to each member
- Member generates x_i and β_i
- Member computes:
- u_i <- g_i^x
- C = h^β * u_i
- PoK{(x): u_i = g_i^x_i)
- Member sends back to dealer:
- Signed {C}
- β_i
- u_i
- PoK{(x): u_i <- g_i^x_i)
- Dealer publishes the commitment on a public bulletin board
-
Dealer's consistency proof:
- Dealer generates a secret vector r_i
- Computes A <- h^r * ∏ u_i^s_i
- Computes e_i from anouncement A using Fiat-Shamir transform
- Computes secrets:
- z = r + \sum_i \beta_i * e_i
- w_i = s_i + e_i
- Announces on public bulletin board:
- Shuffled list of (u_i, g_i, PoK{(x_i): u_i = g_i^x_i}, w_i) and anouncement A, z
-
Verification process:
- Verify that every g_i was generated verifiably random
- Verify PoK for every generator used in the commitment proof PoK{(x): u_i = g_i^x_i}
- Compute e_i from A
- Check that h^z * ∏ u_i^(s_i) = A * ∏ C_i^e_i
The second step can be made less interactive if member instead picks a random seed and generates the generator with it locally and deliver back the seed to the dealer. This scheme is also very convinient to be used for implemetning receipt freeness and coercion resistance that will be discloused shortly in TallyProofs.jl.
To use the protocol we every member creates a commitment with commit method that is provided:
m = g^42
row::CommittedRow, C::G, β::BigItn = commit(m, g::G, h::G, verifier)
where g is unique generator provided to the member by the dealer and m will is a committed message. The commitment is signed by the member and the triple is delivered to the dealer over confidnetial channel. The dealer checks that C = h^β * u, PoK{(x): u <- g^x} and that g is an independent generator provided exlusivelly for the member (publically there is no link between the independent generator and the member's identity).
When the collection period closes the dealer shuffles the resulting commitments and anounces shuffled rows along with a shuffle proof which can be produces via:
simulator = shuffle(rows, h, 𝐂, 𝛃, verifier)
verify(simulator) # true
The essence of the proof is that it checks that every generator is indpendent and that every commitment has a unique generator preventing substitution or removal and that the PoK{(x): u <- g^x} which ensures that u is indepent generator provided that g signs the message m that is passed as suffix to the verifier challenge to the Schnorr proof and hence signs it.
There is a substatntial work done in the literature for automating protocol scheme generation [1, 2]. This is interesting endavour, however, it does sacrifdice the ability to specify resulting protocol whoe would want to implement verification of the proofs independently with the spirit of software independence.
For instance Schor protocol for provong knowledge for a list of tuples (g_i, y_i) can be either written as
PoK{(x_1,...,x_N): ∧_i y_i <- g_i^x_i}
or as
∧_i PoK{x_i: y_i <- g_i^x_i}
proving the same thing. The difference is that in the first case one would use a challenge that lists all inputs in the protocol wheras in the second case every challenge is compputed seperatelly from the inputs.
Choices like theese general compilar makes makes it much harder for one from making an independent verifier for the proofs and their specification. Perhaps one could write a specification for the compiler instead then, but at the present to my knowledge such compilers are unable to produce more complex proofs bulletproofs or proofs of shuffle which need to be added add hoc. So their applicability is limited and one would still need to return to the pen and paper implementation and then implement it. Nevertheless such proof systems can be a valuable tool for finding out and experimenting with new protocol designs.
- [1] Lueks, Wouter et al. zksk: A Library for Composable Zero-Knowledge Proofs. Proceedings of the 18th ACM Workshop on Privacy in the Electronic Society (2019)
- [2] J. B. Almeida, E. Bangerter, M. Barbosa, S. Krenn, A.-R. Sadeghi, T. Schneider. A Certifying Compiler for Zero-Knowledge Proofs of Knowledge Based on Sigma Protocols. In 15th European Symposium on Research in Computer Security (2010)
Other more specialized proofs exist. For instance currently the state of the art range proofs are are made with bulletproofs [3, 4]. In contrast to provided implementation within SigmaProofs their proofs are in logarithmic of the range bitlength wheras here it grows linear with bitrange size. This can be useful for reducoing the size for identification schemes or auctions, but may not be what one needs for homomorphic evoting systems where the ranges are small.
- [3] Bünz, B., Bootle, J., Boneh, D., Poelstra, A., Wuille, P., & Maxwell, G. Bulletproofs: Short Proofs for Confidential Transactions and More. IEEE Symposium on Security and Privacy (2018)
- [4] BulletProofs implementation in Rust https://doc.dalek.rs/bulletproofs/
The SigmaProofs implements Verificatum verifier, parser along with independent generator basis generation that is specified in [5]. The proof of knowledge proofs for logarithm are well explained in lecture notes which I highly recomend anyone starting out their zero knowledge proofs journey [6]. The bit range proofs are implemented according to [7]. The secret sharing is being implemented according to [8]. There is a good overview of zero knowledge proofs used in evoting which one can fairly easy implement with SigmaProofs offered infrastructure [
- [5] Wikstrom, D. How To Implement A Stand-Alone Verifier for the Verificatum Mix-Net.
- [6] Schoenmakers, B. Lecture Notes Cryptographic Protocols (Version 1.9). Department of Mathematics and Computer Science, Technical University of Eindhoven. (2024)
- [7] Ronald Cramer, Rosario Gennaro, and Berry Schoenmakers. A secure and optimally efficient multi-authority election scheme. In Proceedings of the 16th annual international conference on Theory and application of cryptographic techniques (1997) https://link.springer.com/chapter/10.1007/3-540-69053-0_9
- [8] Feldman, P. A practical scheme for non-interactive verifiable secret sharing. In 28th Annual Symposium on Foundations of Computer Science (1987) (pp. 427-438). IEEE. https://www.cs.umd.edu/~gasarch/TOPICS/secretsharing/feldmanVSS.pdf
- [9] Smith, W.D. Cryptography meets voting. (2005)