Okamoto Security Labs · Vortex DFS v0.1

Security that obeys
the laws of physics.
Not probability.

Vortex doesn't predict attacks. It reacts to the laws of exact sciences. A packet either satisfies the mathematics — or it doesn't.

Read the docs View on GitHub
11
Modules
60+
Tests
2
Languages
0
Unsafe in critical paths
Core guarantees

Three laws.
No exceptions.

Every packet is evaluated against deterministic mathematical invariants — not patterns, not models, not predictions.

Post-quantum by design

Signatures based on Learning With Errors — the foundation of NIST's 2024 post-quantum standards. A quantum computer does not break this.

Physics-bound trust

Trust scores derived from distance and entropy metrics, evaluated against deterministic thresholds. Not a model. Not a heuristic. Mathematics.

Zero ambiguity

Every packet is Accept or Reject with a typed, auditable reason. No silent failures. No undefined behavior. No gray area to exploit.

Architecture

Four layers.
One decision.

A packet only reaches the next layer if it fully satisfies the current one.

01
Protocol protocol.rs · main.go
Binary parsing via explicit byte reads. CRC-32 validation. No pointer casting, no unsafe.
02
Authentication vortex_guard.rs
HMAC-SHA256 with constant-time comparison. 1MB body limit. Session IDs sanitized before logging.
03
Signature signer_lwe.rs
Fiat-Shamir over LWE lattice. Hash-bound commitment. Forging requires solving a problem quantum computers cannot.
04
Trust evaluation engine.rs · pqc_core.rs
Distance and entropy evaluated against physics-derived thresholds. Returns a typed TrustBand.
Open source · Apache 2.0

Ready to build on
physics, not probability?

The code is open. Read it, break it, tell us what you find.

View on GitHub Get in touch

Getting started

Vortex DFS is a post-quantum security SDK written in Rust with a Go gateway. It replaces heuristic-based security with deterministic mathematical evaluation — every packet is either valid or it isn't.

Installation

Cargo.toml
[dependencies]
vortex-dfs = "0.1"

Quick start

main.rs
use vortex_dfs::engine::{VortexGate, TrustState};
use vortex_dfs::signer_lwe::keygen;

let (sk, pk) = keygen(seed);
let gate = VortexGate::new(pk.clone());
let sig = sk.sign(&payload, &pk, nonce);

match gate.process_packet(&raw, &sig) {
    TrustState::HighTrust             => { /* proceed */ }
    TrustState::Operational           => { /* monitor */ }
    TrustState::Fragile               => { /* degrade */ }
    TrustState::RejectedSignature     => { /* block */ }
    TrustState::RejectedProtocol(why) => { /* block */ }
    TrustState::RejectedBounds        => { /* block */ }
}

Configuration

terminal
export VORTEX_HMAC_KEY="$(openssl rand -hex 32)"

Trust pipeline

states
├─ CRC mismatch?           → RejectedProtocol
├─ LWE signature invalid?  → RejectedSignature
├─ Metrics out of [0,1]?   → RejectedBounds
└─ score ≥ 0.95            → HighTrust
   score ≥ 0.70            → Operational
   score ≥ 0.20            → Fragile
   score < 0.20            → Critical

Modules

protocol.rs

Binary packet parsing. Uses from_le_bytes() — no pointer casting, no unsafe. CRC-32 IEEE validation.

signer_lwe.rs

Fiat-Shamir over LWE lattice. Hash-bound commitment. For production use pqcrypto-dilithium.

engine.rs

Typestate pipeline with PhantomData. A packet cannot reach evaluation without passing validation — enforced at the type level.

intent_hash.rs

HMAC-SHA256 with constant-time comparison. Invalid hex returns an explicit typed error.

vortex_guard.rs

Axum middleware. 1MB body limit. Session IDs sanitized. HMAC key from environment.

Go gateway

main.go
pkt, err := ParsePacket(buf)
err = VerifySignature(payload, sigHex, key)
raw := BuildPacket(0x0001, payload)

Security properties

  • Post-quantum signatures — Fiat-Shamir over LWE
  • Timing attack resistance — constant-time in Rust and Go
  • No undefined behavior — zero unsafe in critical paths
  • Tamper detection — CRC-32 + hash-bound commitment
  • Log injection prevention — session IDs sanitized
  • DoS prevention — 1MB body limit before allocation
  • No hardcoded secrets — all keys from environment

Production checklist

  • Set VORTEX_HMAC_KEY via secret manager
  • Replace signer_lwe with pqcrypto-dilithium
  • Generate nonces from OsRng
  • Add rate limiting to vortex_guard
  • Run git log -S "SECRET" --all to verify clean history
Okamoto Security Labs · Publications

Research

Technical papers and findings from our work on deterministic security, post-quantum cryptography, and physics-bound trust evaluation.

June 2026CryptographyLWEPost-Quantum
How a tolerance overflow made our post-quantum signatures accept everything

A silent bug in our LWE signature verification caused verify() to return true for any input — including signatures made with completely different keypairs. Root cause: with modulus Q=257, the tolerance c×ETA exceeded Q/2, making the ring distance check vacuously true. We document the discovery, the mathematical analysis, and the fix — moving to hash-bound commitment (Fiat-Shamir). Caught only by explicitly writing the adversarial test case.

May 2026Agentic AIDetection
Detection Fidelity Score: A Framework for Evaluating Detections as Decision Systems

Introducing DFS = S × T × B as a quantitative framework for security detections. Signal Strength evaluates semantic fidelity. Telemetry Stability validates payload completeness under high-throughput runtime. Behavioral Robustness measures resistance to structural drift. Applied to LLM output pipelines as deterministic middleware.

April 2026ArchitectureAI Security
Architectural Mitigations Against Runtime Alignment Drifts in LLM Systems

System prompts alone are insufficient to mitigate runtime alignment drifts — they introduce token overhead, increase latency, and remain susceptible to adversarial injection. A deterministic approach handles this downstream via isolated decision-boundary middleware. By treating LLM output as a raw signal, we compute a Detection Fidelity Score before dispatching the response.

Okamoto Security Labs · Access

Simple, transparent pricing.

One SDK. Pay for the period that fits your deployment cycle. No seats, no per-request fees, no hidden costs.

Starter
$ 9 / week
billed weekly · cancel anytime

Ideal for solo engineers evaluating Vortex DFS in a staging environment.

  • Vortex DFS SDK (Rust + Go)
  • Module 1: Trust & Media
  • Module 2: Code & SecOps
  • Community support (GitHub)
  • Shield Anonymize endpoint
  • Webhook integration
  • SLA & priority support
Enterprise
$ 79 / week
billed weekly · cancel anytime

For organizations with compliance requirements and custom detection rules at scale.

  • Everything in Pro
  • Custom PII pattern registry
  • Dedicated Slack channel
  • 99.9% uptime SLA
  • Priority support · 4h response
  • Security review & onboarding call
  • Invoice billing available
Common questions
Can I switch billing periods? +
Yes. Stripe handles proration automatically — if you upgrade from weekly to annual, you're charged only the difference from your current billing cycle.
What currency do you charge in? +
All prices are in USD. Stripe handles currency conversion at checkout for non-US cards. No FX surcharge from our side.
Is the SDK open source? +
The core Vortex DFS SDK is Apache 2.0 — free to use and audit. The subscription gives you the production modules (/v1/shield/anonymize, webhook handler), K8s manifests, and support.
Can I cancel at any time? +
Yes, any plan can be cancelled from your Stripe customer portal. Access continues until the end of the current billing period.
Do you offer refunds? +
If Vortex DFS doesn't work for your use case within 7 days of purchase, contact us and we'll refund — no questions asked.
7-day money-back guarantee · Payments secured by Stripe · No card stored on our servers
Get in touch

Let's talk
about your stack.

We work with engineering teams evaluating post-quantum migration, AI infrastructure security, and deterministic defense layers. If heuristics haven't solved your problem — reach out.

São Paulo, Brazil · Remote worldwide