Benchmarks
APIsPricingDocsWhite PaperTokenBlogAboutCyber Insurance (HATS)Security Demo
Log InTalk to UsGet API Key
Encrypted Decision Engine

Make decisions on data
you cannot see

Approve, deny, route, and enforce policy — without ever exposing the underlying data. No decryption. No plaintext. No exceptions.

The system evaluates encrypted values and produces a decision. The data is never exposed.

768
encrypted decisions per second · 8-bit comparisons · 96 parallel channels · ARM CPU · no GPU
Get API Key Read Benchmark

Most encryption can compute.
It cannot decide.

Traditional FHE schemes — BFV, CKKS — can add and multiply encrypted data. They are powerful arithmetic engines.

They cannot compare values. They cannot evaluate thresholds. They cannot branch. They cannot enforce rules.

That is why real systems still expose data to make decisions.

TFHE changes that.

Traditional FHE (BFV, CKKS)

Addition. Multiplication. Inner products. Polynomial operations on encrypted data.

TFHE

Comparisons. Equality checks. Threshold decisions. Branching logic. All on encrypted data.

The system cannot see your data

The server does not know the value. It does not know the identity. It does not know the input. It does not know the underlying data.

The system evaluates every possible path and cannot determine which one applies.

Measured Performance

All numbers measured on AWS Graviton4 (ARM), sustained over 30 seconds.

OperationPrecisionThroughput
Comparison8-bit768 TPS
Comparison16-bit372 TPS
Comparison32-bit182 TPS
Comparison64-bit91 TPS
Equality16-bit769 TPS

Every doubling in precision doubles the cost. Performance is linear and predictable.

Automatic routing — no configuration required

You do not choose the encryption scheme. The system does.

Polynomial operations (add, multiply, inner product)
   BFV

Decision operations (compare, branch, match)
   TFHE

The FHE-IQ router selects the correct engine automatically.

What this makes possible

Encrypted fraud detection

Evaluate risk thresholds without seeing transactions. The system compares encrypted scores against encrypted cutoffs and returns a pass or fail.

Encrypted identity matching

Match records across institutions without exposing identities. Encrypted equality checks determine whether two encrypted values correspond — without revealing either one.

Encrypted financial decisions

Approve or reject transactions without revealing amounts. Encrypted comparisons evaluate thresholds on ciphertext, not plaintext.

Encrypted time validation

Check expirations and session validity without reading timestamps. Full 64-bit Unix timestamp comparison on encrypted values at 91 TPS.

What TFHE does not do efficiently.

TFHE is optimized for decision logic: comparisons, equality checks, threshold evaluations, and branching.

It is not efficient for large hash functions, high-depth boolean circuits, or full cryptographic primitives on encrypted data.

Example: SHA3-256 evaluation on encrypted data → approximately 0.30 TPS.

We publish this because performance limits matter.

Post-quantum secure by design

TFHE is based on lattice cryptography (Learning With Errors). The hardness assumption that protects every encrypted decision is the same class of problem that underpins all NIST post-quantum standards.

Every decision is sealed with H33-74 — a 74-byte post-quantum proof. 32 bytes on-chain. 42 bytes off-chain.

Signed using three independent hardness assumptions: MLWE lattices, NTRU lattices, and hash-based signatures.

Every decision is provable

1. Data encrypted on client
2. Decision computed on encrypted data
3. Result reduced to a decision bit
4. Decision signed (post-quantum, three independent hardness assumptions)
5. Proof committed — 74 bytes, permanent

This creates a permanent, verifiable record of a decision made without exposing data.

How it fits together

BFV
High-throughput arithmetic
TFHE
Decision logic
FHE-IQ
Automatic routing
H33-74
Verifiable proof

A system that computes, decides, and proves — without ever exposing data.

Verify It Yourself