Groth16: How the Three-Element zk-SNARK Works

Original paper On the Size of Pairing-Based Non-Interactive Arguments
Author Jens Groth
Publication EUROCRYPT 2016
Official record UCL Discovery: paper record and accepted version

Groth16 is the common name for a preprocessing, pairing-based zk-SNARK construction introduced by Jens Groth in a 2016 paper. It lets a prover show that a public statement follows from some private witness without revealing that witness, provided the statement is encoded in a suitable circuit and the system’s assumptions hold. Groth16 is a proof system, not a blockchain or a complete privacy feature by itself. (UCL Discovery: Groth’s 2016 paper)

What the paper establishes

Groth’s construction proves arithmetic-circuit satisfiability using asymmetric bilinear pairings. A proof contains three elliptic-curve group elements, and verification can be expressed as one pairing-product equation evaluated with three pairings, alongside work involving the public inputs. This compact form helps make proof verification practical in settings where a verifier cannot rerun the full computation. (UCL Discovery: paper abstract and accepted version)

The paper should not be described as proving that three elements are the absolute minimum possible. Its lower-bound result rules out a one-element proof in the stated generic asymmetric-group setting; the authors explicitly leave open whether the argument can be extended to rule out two-element proofs. “Three-element proof” is accurate; “proven smallest possible proof” overstates the result. (Groth, 2016, EUROCRYPT paper)

How a Groth16 proof is produced

  1. Describe the claim. A developer represents a computation as an arithmetic circuit. Public inputs describe the statement that will be checked; the witness contains the private values that satisfy the circuit.
  2. Generate parameters for that circuit. A setup procedure creates a proving key and a verifying key. In the standard Groth16 design, this setup is tied to the circuit. If secret setup randomness were retained or recovered, the soundness guarantees could be compromised. Multi-party ceremonies divide this trust; their security depends on the specific protocol and correct execution.
  3. Create the proof. The prover uses the witness and proving key, together with fresh randomness, to calculate the three proof elements, commonly written as A, B and C.
  4. Verify. A verifier uses the public inputs and verifying key to check a pairing relation. The proof stays small as the circuit grows, although processing public inputs and creating the proof still take resources.

Proof byte size depends on the curve and encoding. For example, the Zcash protocol specification describes a 192-byte Groth16 proof using BLS12-381. That number is an implementation-specific encoding, not a universal byte size for every Groth16 system. The same specification says Sapling uses a modified Groth16 system and its production keys and circuits; this is an application of the construction, not an exact copy of Groth’s original paper. (Zcash Protocol Specification, Groth16 and proof encoding)

Where it is used

Zcash’s Sapling protocol uses a Groth16-based proving system for shielded transaction statements. Ethereum’s alt_bn128 pairing precompile, specified by EIP-197, provides a building block that smart contracts can use for pairing checks in SNARK verification. These examples do not mean that every privacy transaction or ZK application uses Groth16; each application defines its own circuit, keys and public statement. (Zcash Protocol Specification; Ethereum EIP-197)

Trade-offs and limits

Groth16 is attractive when compact proofs and efficient verification matter. Its main operational cost is circuit-specific setup: changing the circuit generally requires new parameters and a new ceremony. Proof generation may also be computationally intensive. Systems such as PLONK explore different setup and proof trade-offs; they are alternatives, not a simple replacement that makes Groth16 obsolete. The original paper’s security analysis is also tied to its cryptographic model and assumptions.

Finally, zero knowledge does not hide information that an application makes public, fix an incorrect circuit, or protect a user’s device and network metadata. A Groth16 proof can attest only to the statement represented by the circuit and the inputs supplied to it. Application design and implementation remain part of the security model.

Related Terms

  • zk-SNARK — the broader class of succinct non-interactive arguments to which Groth16 belongs.
  • Zero-Knowledge Proof — the privacy property that a proof can establish without disclosing the witness.
  • zkML — one area that uses proofs to verify computations, with different systems chosen for different workloads.

See Also

  • Zcash Sapling MPC — a paper on multi-party generation of zk-SNARK parameters.
  • PLONK — a later proof system with a different setup model and design trade-offs.
  • Zerocash — an earlier privacy-payment design that applied zk-SNARKs.

Last reviewed: October 28, 2026.

Sources