TorusDB
The encrypted data platform. Store, query, and serve data that is never decrypted. Not at rest, not in transit, not even in use.
Built to run where encryption usually can’t.
Not an FHE that does everything. Just what a database actually needs, on air-gapped CPUs, today.
Standard CPUs only
No GPU or TPU cluster required. It drops straight into the air-gapped networks of banks, government, and electronics makers where GPU-based encryption can’t go.
No schema changes
Column types are preserved and no ALTER TABLE is needed. It applies instantly to existing Oracle, MSSQL, SAP HANA, PostgreSQL, and MySQL.
Real-time core DB ops
Comparison (WHERE) and aggregation (SUM) are the operations databases lean on most. They run on ciphertext in microseconds.
The problem it solves
Data has to be decrypted to be used. So every query, analytics job, and breach becomes a moment of plaintext exposure for attackers and insiders.
Store and query data that is never decrypted. The plaintext-exposure window simply doesn't exist, at rest, in transit, or in use.
Encrypted across the entire lifecycle
Queries run directly on ciphertext, so there is no moment of plaintext exposure for an attacker, or an insider, to exploit.
Tweezer encryption: only what matters
Apply encryption like tweezers, to just the highest-risk columns (national ID, account number, transaction memo). Keep your DB architecture and isolate the data that carries the risk.
Fits your existing data stack
Works alongside the systems of record you already run, so teams adopt privacy-preserving data without re-platforming.
How TorusDB works
Encryption stays on the data through its whole life, from the moment it's written to the moment a result is read.
Encrypt at the source
Sensitive columns are encrypted before they ever land, so plaintext never touches storage.
Store as ciphertext
Data sits encrypted at rest, and keys are never pooled in one place to steal.
Query on ciphertext
SQL runs directly on encrypted data: filtered, joined, and aggregated without decryption.
Return encrypted results
Only the entitled client decrypts, at the edge. The platform never sees plaintext.
Wondering if this fits your environment?
Take 15 minutes with a security engineer, against your real workload.
Related use cases
See where this runs in production, across AI, finance, and core enterprise systems.
Frequently asked
Quick answers about this product.
We don't claim quantum resistance as a mathematical property of the cipher itself. Instead, TorusDB's No-Key architecture removes what a quantum attack would target: no public, evaluation, or secret key ever lives on the server, so there is nothing there for a Shor-style attack to derive. The server's view is designed to be information-theoretically independent of the plaintext, so even a quantum adversary has no key to break and no readable data to take. For quantum-resistant signing and authentication, PrivID adds a lattice-based layer (CRYSTALS-Dilithium, the NIST FIPS 204 standard) on top.
Yes. Homomorphically encrypted data doesn't reveal the size, order, or distribution of the plaintext, so a plain B-tree index no longer applies. TorusDB instead provides a token-based Secure Index that filters quickly while data stays as ciphertext, and uses parallelism to accelerate heavy queries.
Throughput scales with core count and parallelism, from single-core encryption to multi-node pipelines, and is tuned to your environment. Concrete figures for environments like ERP and SAP HANA are measured and shared through a PoC.
No. The team running analysis holds neither the plaintext nor a decryption key. Statistics, aggregation, conditional search, and segmentation run on ciphertext, and results are produced as ciphertext by default. When a final decryption is genuinely required, no individual holds the key alone. It's handled selectively via threshold key sharing / MPC / policy-based approval.
Column-level keys, role-based access, and a KMS are security policies: they govern who may access data. Homomorphic encryption is a compute technology that runs operations on ciphertext without decrypting it. They aren't competitors but complements. TorusDB's goal is to run statistics, aggregation, matching, and analysis without ever exposing plaintext.
See TorusDB on your data.
Bring a real workload. We'll show queries running on encrypted data.