Product 03 — Active Research

iZKP Verifier

An interactive Zero-Knowledge Proof system with an AI Small Language Model verifier — enabling users to prove their identity through a cryptographic challenge-response exchange without ever revealing their credentials.

Inquire about research View all products
Interaction model
Challenge-response between prover & verifier
Protocol family
Sigma (Σ) Protocol — three-phase cryptographic exchange
Research status
Active R&D — MyKey, Inc.

Overview

Prove it.
Without revealing it.

An interactive Zero-Knowledge Proof (iZKP) is a cryptographic approach that enables one party — the prover — to prove to another party — the verifier — that they know a piece of information, such as a password or credential, without revealing that information itself.

MyKey's iZKP system pairs this with an AI Small Language Model (SLM) as the verifier, embedded directly on-device. The result is a novel authentication architecture — iZKP-AIV — where the AI verifier engages in a dynamic, iterative challenge-response dialogue with the prover, gaining cryptographic confidence in the user's identity without ever seeing or storing the underlying secret. There is currently no known practical, production-ready implementation of this architecture anywhere in the world.

The cryptographic foundation

The Sigma (Σ) Protocol.
Three phases. Zero exposure.

iZKP-AIV is built on a family of three-step cryptographic protocols known as Σ-protocols (Sigma Protocols). The Greek letter Σ visualizes the flow of messages between the prover and verifier. Each phase is carefully designed so that the AI verifier gains confidence in the user's identity while learning nothing about the secret itself.

Phase 1
Commitment
Prover → Verifier

The prover generates a random nonce and uses it to create a cryptographic commitment — a binding, non-interactive promise sent to the AI verifier. This commitment conceals the prover's secret while locking in their position before the challenge is issued.

Secret remains hidden
Phase 2
Challenge
Verifier → Prover

The AI verifier sends a random challenge value to the prover. This challenge must be unpredictable and issued only after the commitment is received — preventing any pre-computation or manipulation by a dishonest prover. The AI's adaptive logic dynamically shapes each challenge.

AI-generated, randomized
Phase 3
Response
Prover → Verifier

The prover computes a response using their secret witness, the original nonce, and the verifier's challenge. The AI verifier then evaluates the commitment, the challenge, and the response together — confirming the prover's identity without ever seeing the underlying credential. Homomorphic encryption ensures the credential elements are never exposed to the SLM verifier at any point.

Credential never exposed

How iZKP-AIV works

Authentication without exposure.

01
User initiates authentication
The user attempts to authenticate on a device with the iZKP-AIV system embedded. Rather than transmitting a password or biometric, the system initiates a Sigma Protocol exchange between the user (prover) and the on-device AI SLM verifier.
02
Prover sends commitment
The prover generates a random nonce and produces a cryptographic commitment — sent to the AI verifier as a binding, non-revealing promise. The user's actual credential never leaves the device in any form the verifier can read.
03
AI verifier issues challenge
The AI SLM verifier generates a randomized challenge value and sends it to the prover. The AI's adaptive logic means the challenge is dynamically shaped based on the session — making pre-computation or replay attacks computationally infeasible.
04
Prover responds, verifier confirms
The prover computes a response using their secret, the nonce, and the challenge. The AI verifier evaluates all three together — wrapped in homomorphic encryption — and either confirms or rejects the authentication. At no point is any credential stored, transmitted in plaintext, or exposed to the verifier.

Applications

Where iZKP-AIV changes everything.

🏛
Government credentialing
Secure credentialing across DHS and federal systems — users prove authorization through a challenge-response protocol without revealing identity data, clearance levels, or biometric templates.
🔐
Zero-trust access control
Replace password and token-based authentication with cryptographic proofs that cannot be stolen, replayed, or forged — directly compatible with zero-trust architecture frameworks.
🤖
AI-enhanced verification
The embedded SLM verifier uses adaptive AI logic to analyze proof patterns, detect anomalies, and identify potential deception attempts that rule-based systems would miss entirely.
🏦
Financial compliance
Prove KYC/AML compliance, accredited investor status, or regulatory clearance through interactive proofs — without disclosing the underlying personal or financial data to any counterparty.
🧬
Biometric integration
iZKP-AIV can be combined with biometric authentication — allowing users to prove identity based on biometric data without ever transmitting or storing the raw biometric template.
🖥
On-device authentication
The SLM verifier is embedded directly on the device — eliminating reliance on centralized credential servers and dramatically reducing the impact of data breaches on the authentication system.

Cryptographic guarantees

Three properties.
No exceptions.

Completeness
An honest prover with valid credentials will always convince the AI verifier. If the user possesses the correct credentials and follows the protocol, iZKP-AIV will successfully authenticate them — every time.
Soundness
A dishonest prover cannot convince the AI verifier of a false claim. The randomized challenge and AI-adaptive verification ensure an attacker without valid credentials cannot falsely authenticate — even with sophisticated forgery attempts.
Zero-knowledge
The AI verifier learns nothing about the user's secret credentials beyond the fact that they are correct. Homomorphic encryption ensures credential elements are never exposed to the SLM at any stage of the interaction.