What Are Zero-Knowledge Proofs? | Forkast Learn

How zero-knowledge proofs work

Every zero-knowledge proof involves three properties that must hold simultaneously:

Completeness. If the statement is true and both parties follow the protocol, the verifier will be convinced. A true statement can always be proven.

Soundness. If the statement is false, no cheating prover can convince the verifier except with negligible probability. A false statement cannot be faked into looking true.

Zero-knowledge. The verifier learns nothing beyond the fact that the statement is true. The proof reveals no information about the underlying secret—no bits, no partial hints, no structural clues [1].

These three properties together create a powerful primitive: verifiable truth without disclosure.

Interactive vs. non-interactive

The original 1985 formulation required back-and-forth communication between prover and verifier—multiple rounds of challenge and response. This is called an interactive proof. In practice, most modern systems use non-interactive proofs, where the prover generates a single proof object that any verifier can check without further communication. This shift, enabled by the Fiat-Shamir heuristic (converting interactive protocols into non-interactive ones using hash functions), made zero-knowledge proofs practical for blockchain and distributed systems [2].

The two main proof systems: SNARKs vs. STARKs

The two dominant families of zero-knowledge proof systems differ in fundamental tradeoffs:

ZK-SNARKs (Zero-Knowledge Succinct Non-Interactive Argument of Knowledge) produce small proofs—typically hundreds of bytes—that verify in milliseconds. This makes them cheap to verify on-chain. However, most SNARK constructions require a trusted setup ceremony: a one-time generation of public parameters where the randomness must be destroyed afterward. If the setup randomness is compromised, the soundness guarantee breaks. SNARKs also rely on elliptic curve cryptography, making them vulnerable to quantum computing advances [3].

ZK-STARKs (Zero-Knowledge Scalable Transparent Argument of Knowledge) require no trusted setup—they use public randomness, making them fully transparent. They are also quantum-resistant because they rely on collision-resistant hash functions rather than elliptic curves. The tradeoff: STARK proofs are larger (tens to hundreds of kilobytes), which increases on-chain verification cost. However, STARKs scale better for large computations—their proving time grows quasi-linearly rather than super-linearly [3].

The practical choice: Applications where on-chain verification cost dominates (small-scale DeFi, identity proofs) tend to favor SNARKs. Applications where transparency, quantum resistance, and large-computation scalability matter (high-throughput rollups, enterprise systems) increasingly favor STARKs.

First production deployment: Zcash

Zcash, launched October 28, 2016 by Electric Coin Co., was the first blockchain to deploy zk-SNARKs in production. It enabled “shielded transactions” where the sender, receiver, and amount are hidden from the public blockchain while still being mathematically verified as valid. This demonstrated that zero-knowledge proofs could move from academic theory to production-grade financial infrastructure [4].

ZK-rollups: scaling Ethereum

The largest production use of zero-knowledge proofs today is ZK-rollups—Layer 2 scaling solutions that move computation off Ethereum’s mainnet, batch thousands of transactions, generate a validity proof, and post that proof back to Layer 1 for settlement. Ethereum’s mainnet verifies the proof in a single check, confirming the entire batch without re-executing each transaction [3].

Notable ZK-rollup projects in production (as of 2025-2026) include:

• Polygon zkEVM — EVM-equivalent ZK-rollup
• zkSync Era — EVM-compatible with native account abstraction
• StarkNet (StarkWare) — STARK-based, Cairo-native execution environment
• Scroll — bytecode-level EVM equivalence
• Linea (Consensys) — Type 2 zkEVM
• Taiko — Based (decentralized sequencer) ZK-rollup

The ecosystem is converging on zkEVM as the dominant pattern: ZK-rollups that recreate EVM opcodes inside ZK circuits, allowing developers to deploy standard Solidity smart contracts while still producing validity proofs. The direction of travel is toward multi-prover architectures, blob-backed data availability (EIP-4844), progressive sequencer decentralization, and cross-rollup proof aggregation [3].

Zero-knowledge proofs and agent identity

Zero-knowledge proofs are increasingly relevant to AI agent identity and privacy-preserving compliance. Microsoft Research’s Vega system (2024-2025) enables AI agents to prove facts about the human they represent—age, personhood, professional status—without the agent ever holding the underlying credential. Vega uses device binding and generates a ~2 KB proof in approximately 92 milliseconds on commodity mobile devices [5].

Google Wallet has integrated ZKP-based selective disclosure so users can prove attributes (e.g., “I am over 18”) without revealing or enabling tracking of the underlying identity document. This is the pattern that agent identity systems are converging toward: cryptographic verification of attributes without exposure of the underlying data [5].

The ZK-KYC market—privacy-preserving compliance for know-your-customer and anti-money-laundering requirements—was projected at approximately $83.6 million in 2025, growing to an estimated $903.5 million by 2032, as institutions adopt ZKPs for regulatory compliance without mass data disclosure [9].

Standardization efforts

ZKProof.org is the primary community-driven standardization initiative. It operates through annual workshops (the 8th workshop is scheduled for May 2026 in Rome) and produces the ZKProof Community Reference document, which compiles terminology, explanations, and recommendations for implementing ZKP schemes and protocols. The goal is interoperability and security standards across implementations [6].

NIST has not published a dedicated standalone ZKP standard, but zero-knowledge proof-based schemes were evaluated as candidates in the Post-Quantum Cryptography standardization process (2017-2024). The finalized PQC standards (FIPS 203, 204, 205, released August 2024) focused on lattice-based and hash-based schemes rather than ZKP-based approaches, though NIST continues to evaluate additional algorithms [7][8].

Limitations and open challenges

Trusted setup risk. Many SNARK systems require a ceremony where participants generate and then destroy secret randomness. If even one participant is honest, the setup is secure—but this assumption is non-trivial. Newer systems (PLONK, Halo 2) reduce but do not eliminate this dependency.

Prover computational cost. Generating proofs is computationally expensive—often orders of magnitude more expensive than executing the computation directly. Hardware acceleration (GPUs, FPGAs, ASICs) is an active area of development.

Proof size and on-chain cost. STARK proofs are large, increasing gas costs for on-chain verification. Recursive proof aggregation (proving proofs of proofs) and proof compression are active research areas.

Developer complexity. Writing ZK circuits requires specialized knowledge different from standard programming. zkVMs (virtual machines that abstract circuit design) are lowering this barrier, but the tooling ecosystem is still maturing.

No standard. As of mid-2026, no single ZKP standard has been ratified. The ecosystem relies on implementation-specific parameters, creating interoperability friction.

Sources

[1] Goldwasser, Micali & Rackoff. “The Knowledge Complexity of Interactive Proof-Systems.” ACM STOC ’85. PDF

[2] Ethereum Foundation. “Zero-knowledge proofs.” Ethereum.org, 2024. Link

[3] Ethereum Foundation. “Zero-knowledge rollups.” Ethereum.org, 2024. Link

[4] Electric Coin Co. “Zcash Launch Announcement.” October 28, 2016. Link

[5] Microsoft Research. “Vega: AI agents can now prove facts about you—without seeing your credentials.” 2024. Link

[6] ZKProof.org. “ZKProof Community Reference.” 2026. Link

[7] NIST. “Post-Quantum Cryptography Standardization.” August 2024. Link

[8] NIST. “Post-Quantum Cryptography: Selected Algorithms.” August 2024. Link

[9] Stratistics MRC. “Zero-Knowledge KYC Market Forecasts to 2032.” 2025. Link