PhylaxOnePhylax ZeroTrust
Authentication Protocol

Phylax ZeroTrust

Distributed post-quantum authentication for the systems that cannot fail.

No secrets stored anywhere. No single point of failure. Native AI agent governance. Built on FROST threshold signing and NIST FIPS 203/204/205 cryptography. Certifiable for CMMC, FedRAMP 20x, and NSS acquisitions from deployment.

NIST FIPS 203/204/205FedRAMP 20x KSI-IAMIETF RFC 9591 (FROST)CMMC Level 2NSA CNSA 2.0

The Problem

Every existing auth system stores something.

Password hashes. Session tokens. Certificate stores. Private keys in hardware security modules. The format changes. The fundamental problem does not: there is a stored secret, and whoever controls that secret controls access.

Classical zero trust frameworks addressed access policy. They did not address the stored credential problem. Your identity provider still holds a database of secrets. Your PKI infrastructure still holds certificate private keys. Breaching those systems -- or the humans who manage them -- still produces access.

Phylax ZeroTrust does not reduce the risk of a credential store breach. It eliminates the credential store.

How It Works

The Signing Ceremony

Based on FROST (Flexible Round-Optimized Schnorr Threshold Signatures, IETF RFC 9591). Two rounds. No stored secrets. No vote. A cryptographic proof or nothing.

01

Entity requests access

A person, organization, AI agent, or device submits an authentication request to a protected resource. The request identifies the entity by its Decentralized Identifier.

02

Node network receives the request

The coordinator service distributes the request to the N participating nodes. Each node holds a key share. No node holds the complete key -- it does not exist as an assembled value.

03

Round 1: commitment generation

Each participating node generates a fresh nonce and publishes a cryptographic commitment. Commitments are collected by the coordinator and distributed to all participating nodes.

04

Round 2: partial signature computation

Each node independently computes a partial signature using the request message, the collected commitments, and its own key share. Partial signatures are returned to the coordinator.

05

Threshold combination and verification

The coordinator combines K or more valid partial signatures into a single cryptographic proof. Either the combined signature is valid or it is not -- there is no vote, no counting, no pass/fail per node. The resource verifies the proof against the public key and applies the Verifiable Credential scope to determine what the entity is authorized to do.

Capabilities

What It Does

No Secrets Stored

No authentication credential exists as a stored value on any node. Each node holds a mathematical key share. Compromise any number of nodes below the threshold and you have zero information about the key or any identity in the system.

Post-Quantum Cryptography

ML-KEM (FIPS 203) for all key encapsulation. ML-DSA (FIPS 204) for all credential signatures. AES-256 throughout. All three are on the NIST approved algorithm list. Certifiable for NSS acquisitions under CNSA 2.0 from day one.

Identity for Every Entity Type

People, organizations, AI agents, IoT devices, and services all hold W3C Decentralized Identifiers. Authentication answers who is this entity and what is it authorized to do -- not just whether a password matches.

Native AI Agent Governance

Each AI agent receives a DID at instantiation. A scoped Verifiable Credential defines exactly which systems it can access, which actions it can take, and when that authorization expires. High-stakes operations require threshold human approval before the signing ceremony completes. Every agent action writes to an append-only signed log.

Adaptive Behavioral Intelligence

Every authentication event builds a per-identity behavioral baseline. Deviations trigger automatic threshold increases or step-up authentication. The K parameter in the signing ceremony adjusts in real time based on risk score. The system does not just respond to attacks -- it gets measurably harder to penetrate the longer it runs.

Fault Tolerant by Design

In a 4-of-7 deployment, any three nodes can be simultaneously offline, under attack, or compromised. The signing ceremony continues with the remaining nodes. There is no manual repair process, no service interruption, and no single administrator who can restore a failed node that others must trust.

Comparison

Classical Zero Trust vs. Phylax ZeroTrust

Classical Zero Trust

Credentials stored on servers

Phylax ZeroTrust

No credential stored anywhere

Classical Zero Trust

Pass/fail vote counts agreeing nodes

Phylax ZeroTrust

FROST threshold ceremony -- no vote, cryptographic proof only

Classical Zero Trust

Fixed authentication threshold

Phylax ZeroTrust

Adaptive K threshold adjusts to real-time risk score

Classical Zero Trust

No standard for AI agent identity

Phylax ZeroTrust

Native DID + scoped Verifiable Credentials + threshold human approval

Classical Zero Trust

Manual repair when nodes fail

Phylax ZeroTrust

Protocol continues through node failures automatically

Classical Zero Trust

Rules-based anomaly detection

Phylax ZeroTrust

Per-identity behavioral baseline with adaptive response

Classical Zero Trust

Point-in-time compliance assessment

Phylax ZeroTrust

Continuous append-only signed audit log

Classical Zero Trust

AES-128 or RSA at the core

Phylax ZeroTrust

ML-KEM + ML-DSA + AES-256 (NIST FIPS 203/204/205)

Who It's For

Built for Specific Environments

Federal and Defense

CMMC Level 2 evidence and audit trail from day one
FedRAMP 20x KSI-IAM alignment -- native, not retrofitted
NSS acquisitions under CNSA 2.0 certifiable at deployment
AI agent governance satisfying OMB M-24-10 requirements
Append-only signed log for all authentication events

Financial Institutions

Correspondent banking auth with no single point of failure
CNSA 2.0 quantum-readiness before the January 2027 deadline
Treasury and high-value transaction authorization
Node distribution across geographic regions -- no regional outage affects authentication
Behavioral baseline detects coordinated fraud patterns across accounts

Enterprise Security

AI agent governance across multi-agent deployments
Dynamic threshold adjusts to risk level per request -- no static rules to maintain
Works alongside existing Active Directory and LDAP deployments
Continuous behavioral baseline replaces annual compliance assessments
Client SDK for JavaScript, Python, and Go -- integration without protocol expertise

FAQ

Hard Questions

How is this different from our existing zero trust product?

Most zero trust products -- including enterprise identity platforms from major vendors -- authenticate by checking a stored credential. A password hash, a certificate, a session token. That stored value is the target in a breach. Phylax ZeroTrust stores nothing. Each node holds a mathematical share that has no value without the others. There is no credential database to breach, no certificate store to rotate after a compromise, and no password reset process to social-engineer.

What happens when a node goes offline or is compromised?

The signing ceremony continues with the remaining healthy nodes. In a 4-of-7 deployment, any three nodes can be simultaneously offline, under attack, or confirmed compromised -- authentication continues at full capacity. The affected node is excluded automatically. When it comes back online, it resynchronizes without any manual intervention or emergency key rotation.

Can this work alongside our existing Active Directory or LDAP deployment?

Yes. Phylax ZeroTrust issues Decentralized Identifiers to entities that already exist in your identity systems. It does not require replacing existing infrastructure. Existing Active Directory groups and LDAP schemas map to Verifiable Credential scopes during a phased migration. The two systems operate in parallel until migration is complete.

How does AI agent governance work in practice?

Each AI agent receives a DID at instantiation. A Verifiable Credential issued against that DID defines which systems the agent can access, which actions it is permitted to take, which data classifications it can touch, and when that authorization expires. For operations above a defined sensitivity threshold, the signing ceremony requires agreement from a set of human approver nodes before completing -- the agent cannot self-authorize beyond its credential scope. Every action the agent takes is attributed to its DID and written to an append-only signed log that satisfies FedRAMP 20x KSI-IAM and OMB M-24-10 audit requirements.

Ready to go deeper?

All briefings are conducted under NDA. Bring your security architecture questions. We will bring the protocol specification and compliance documentation.

Accepting qualified partner inquiries