# Post-Quantum Attribute-Based Encryption: Research Status

**Source**: https://quantumsequrity.com/blog/pq-attribute-based-encryption
**Category**: Future Research

---

[← Back to Blog](../../blog) Future Research

# Post-Quantum Attribute-Based Encryption: Research Status

11 min read

Attribute-Based Encryption (ABE) lets you encrypt files so that anyone whose attributes match a policy can decrypt, without listing the recipients explicitly. Encrypt a personnel record under "department:HR AND clearance:secret AND tenure>2years", and any HR employee with secret clearance and more than two years of tenure can read it. The keyholder set is determined at decryption time, by who has the right credentials, not at encryption time. ABE is one of the more powerful generalisations of public-key encryption and is the natural fit for cloud storage where the recipient set is not known in advance.

Most ABE deployments today use pairing-based math. A sufficiently large quantum computer would break them. Post-quantum ABE has been an active research area since the late 2000s, with lattice-based constructions reaching practical efficiency only recently. This article walks through what ABE does, the two flavours (KP-ABE and CP-ABE), and where post-quantum variants stand.

## What ABE Solves

Imagine you run a hospital. You have 10,000 employees, and a sensitive patient record needs to be encrypted so that:

- Any oncologist on staff can read it.
- Any nurse in the oncology ward can read it.
- The patient's primary care physician can read it.
- Any administrator with audit rights can read it.

With standard public-key encryption, you would have to encrypt the file 50 to 200 times, once for each recipient. With ABE, you encrypt once with a policy: "(role:oncologist) OR (role:nurse AND ward:oncology) OR (relation:primary_physician) OR (role:admin AND scope:audit)". Any employee whose attributes satisfy that boolean expression can decrypt. New employees who later acquire matching attributes can also decrypt. Old employees who lose attributes lose access (with caveats around revocation).

Two flavours:

**Key-Policy ABE (KP-ABE)**: The policy is in the user's key. The ciphertext is tagged with attributes. Decryption succeeds if the user's policy is satisfied by the ciphertext's attributes. Useful when the data publisher is known in advance and the policy varies by user.

**Ciphertext-Policy ABE (CP-ABE)**: The policy is in the ciphertext. The user's key is tagged with attributes. Decryption succeeds if the user's attributes satisfy the ciphertext's policy. Useful when the data publisher decides who can read at encryption time.

CP-ABE is more common in practice because it matches the natural workflow of "I encrypt, I specify the policy, anyone with matching attributes decrypts".

## Why Post-Quantum ABE Is Hard

Classical ABE relies on bilinear pairings, specifically a function `e(g^a, g^b) = e(g, g)^(ab)` where the equation works because pairings have a specific algebraic structure. The structure lets you build sophisticated access policies that decrypt only when the right combination of attributes is present.

Lattices do not have pairings. To build ABE on lattices, researchers had to find different techniques. The breakthrough came from Gorbunov-Vaikuntanathan-Wee in 2013, who showed that lattice ABE could be constructed using "matrix branching programs" or general circuits. The construction works but is much less efficient than pairing-based ABE.

Three challenges in post-quantum ABE:

1. **Ciphertext size**. Lattice ABE ciphertexts are typically 10x to 100x larger than pairing ABE ciphertexts.
2. **Decryption time**. Decryption can be slow because the user has to evaluate a circuit-like structure.
3. **Policy complexity**. Some lattice ABE schemes only support simple policies (AND, OR with limited depth). Others support general circuits but at very high cost.

For lattice background, see [Lattice-Based Cryptography Explained](lattice-based-cryptography-explained).

## Lattice ABE Constructions

Several lattice-based ABE schemes have been published:

**Boyen ABE (2013)**: One of the first lattice CP-ABE constructions. Supports threshold policies (at least k of n attributes). Reasonable parameters but limited expressiveness.

**Gorbunov-Vaikuntanathan-Wee (2013)**: Lattice KP-ABE for general circuits. Theoretical breakthrough but parameters too large for most practical use.

**Boneh-Gentry-Gorbunov-Halevi-Nikolaenko-Segev-Vaikuntanathan-Vinayagamurthy (BGG+ 2014)**: KP-ABE that scales to circuit-style policies with smaller parameters than GVW. Foundation for several follow-up schemes.

**Brakerski-Vaikuntanathan (BV 2016)**: A more practical lattice ABE with sub-linear decryption time for some policy classes.

**Datta-Pal-Yamada (DPY 2021)**: A lattice CP-ABE with smaller ciphertexts and decryption keys than previous schemes. Supports a useful subset of policies.

**Agrawal-Yamada (AY 2020)**: Lattice ABE with broadcast and revocation features. Parameters are larger but the feature set matches some real cloud workflows.

For the underlying ML-KEM-style structure, see [ML-KEM Explained](ml-kem-explained).

## Production ABE: What Exists Today

Several ABE products are deployed:

**Zeutro**: Cloud encryption product using pairing-based CP-ABE. Migration path to lattice ABE is on the published roadmap.

**Cipher Cloud / Lookout**: Some enterprise data-at-rest encryption uses ABE. Mostly pairing-based.

**Healthcare and government cloud**: Pilot deployments of ABE for medical record sharing and inter-agency document exchange. Mostly pairing-based.

**Open-source libraries**: openABE (formerly Voltage's library) supports several pairing-based ABE schemes. No lattice ABE library has reached comparable production maturity as of 2026.

The barrier to lattice ABE in production is not just standardisation. It is also performance. A 100x larger ciphertext is a real bandwidth cost, and a 10x slower decryption is a real latency cost. Recent work is closing the gap, but ABE on lattices is still slower than ABE on pairings.

## Cloud Encryption and ABE

The natural use case for ABE is cloud storage where:

- The data owner does not know all future recipients at upload time.
- Recipients are determined by their organisation-managed attributes.
- The cloud provider should not be trusted with plaintext.

This is the model some healthcare and finance cloud platforms use. The data owner encrypts under a policy, the cloud stores the ciphertext, and any user whose attributes match (verified through key issuance, not through cloud trust) can decrypt.

QNSQY's main encryption model is different: the data owner names the recipients at encryption time and encrypts one file to all of them. QNSQY also has attribute-based modes, described below.

## Revocation: The Hard Problem

A practical challenge in ABE is revoking attributes. If an employee leaves the company, their "employee" attribute should no longer let them decrypt. Two approaches:

**Time-bound attributes**: Attributes have an expiry. Users get a new credential each period. Decryption keys for past periods still decrypt past ciphertexts (which is sometimes desired, sometimes not).

**Direct revocation**: The KGC publishes a revocation list. Decryption requires checking the list. Lattice schemes for direct revocation exist but add complexity.

Both approaches add overhead. For ABE in cloud workflows where employees frequently change roles, the revocation cost can dominate the benefit. This is a deployment consideration, not specific to post-quantum.

## Functional Encryption: ABE's Big Brother

Functional Encryption (FE) generalises ABE: instead of "decrypt if policy matches", FE allows "compute function f on the encrypted message and output f(message), nothing else". ABE is a special case of FE where f is the identity function plus a policy gate. See [PQ Functional Encryption](pq-functional-encryption) for the broader picture.

Lattice FE is even more research-grade than lattice ABE. Most schemes are theoretical with parameters too large for practical use. ABE is the practical subset that has any chance of production in the next few years.

## Multi-Authority ABE: Beyond a Single KGC

In single-authority ABE, one KGC issues all attribute keys. This is a bottleneck and a single point of trust. Multi-authority ABE (MA-ABE), introduced by Chase in 2007, allows multiple authorities each managing different attributes. A user assembles their key from multiple authorities, and decryption succeeds when the policy is satisfied across the combined attributes.

Use case: a hospital network where the medical college issues "specialty" attributes, the hospital HR issues "department" attributes, and the credentialing body issues "license" attributes. A doctor's effective access depends on attributes from all three authorities.

Lattice MA-ABE constructions exist (Rouselakis-Waters extensions, Lewko-Waters-style adaptations) but with higher overhead than single-authority ABE. The added complexity is worth it for federated environments where no single party should control all access.

## Decentralised ABE

A more recent direction is decentralised ABE (DABE), where there is no central KGC at all. Authorities can be added or removed without re-keying the system, and users can mix-and-match authorities. Lattice DABE constructions are in early research, with the most cited being Lewko-Waters DABE adapted to lattices. Performance is significantly slower than centralised ABE.

For organisations that need ABE-style access control but cannot tolerate any single trusted authority, DABE is the long-term direction. As of 2026, no production DABE deployment exists in any cryptographic family.

For comparison with multi-recipient encryption, see [Hybrid Encryption](hybrid-encryption).

## What QNSQY Does and Does Not Cover

QNSQY has two attribute-based modes. On Pro and Business, `qnsqy abe encrypt` encrypts a file to every listed recipient whose attributes satisfy a policy such as `(role=doctor AND dept=cardio) OR clearance>=4`; the policy is evaluated at encryption time, so the result is a multi-recipient file for a known set of recipients. On Business, an attribute authority issues users bundles of per-attribute post-quantum KEM keys, and `qnsqy abe seal` encrypts a file under a monotone policy so that only bundles whose attributes satisfy it can reconstruct the key. Both are built from post-quantum KEMs and secret sharing, not from one of the lattice ABE schemes described above.

For users who need lattice-based ABE specifically, the recommendation is:

- Use a research-grade lattice ABE library if your application can accept the size and speed costs.
- Wait for NIST or IETF standardisation before deploying in compliance-bound systems.
- Consider whether multi-recipient encryption (encrypting once to a fixed list of known recipients) meets your needs.

QNSQY's model is appropriate when recipients are known at encryption time. The QNSQY multi-recipient feature (Pro and Business) lets you encrypt once to N pre-listed recipients in a single file, which solves a practical subset of the ABE use case without requiring full attribute-based machinery. For organisations whose access policies are stable enough to be expressed as recipient lists rather than attribute predicates, multi-recipient encryption is simpler and faster.

For organisations with more complex policies (boolean expressions over many attributes, fine-grained role-based access), waiting for standardised lattice ABE remains the right call.

The practical pattern for most cloud workloads today is: encrypt at-rest data with ML-KEM hybrid encryption (which QNSQY does), and enforce access control at the application layer using identity and attribute systems (Active Directory, Okta, Auth0). The application layer makes the access decision, then either decrypts the file (if access is granted) or returns nothing. This is not cryptographic ABE, since the application could be compromised, but it is the practical compromise most organisations make today while waiting for ABE to mature.

## Frequently Asked Questions

### Is any post-quantum ABE standardised?

No. As of 2026, no NIST or IETF standard exists for post-quantum ABE. Several lattice ABE schemes are well-studied research candidates, but none has gone through standardisation.

### What is the difference between KP-ABE and CP-ABE?

In KP-ABE, the policy is in the user's key, and the ciphertext is tagged with attributes. In CP-ABE, the policy is in the ciphertext, and the user's key is tagged with attributes. CP-ABE is more common because it matches the natural workflow where the data publisher decides the policy at encryption time.

### How big is a lattice ABE ciphertext?

Depends on the policy and the scheme. For typical policies (a few attributes, simple boolean structure), lattice ABE ciphertexts are 10 KB to 100 KB. Pairing-based ABE for the same policies is around 1 KB. The size penalty is one of the main barriers to lattice ABE in production.

### Can I revoke an attribute in ABE?

Yes, but it is more complex than revoking a regular public key. Two approaches: time-bound attributes (where users get fresh keys each period) and direct revocation (where the KGC publishes a list of revoked users). Both add overhead. Lattice schemes support both with similar complexity.

### Is ABE related to identity-based encryption?

Yes. IBE is a special case of ABE where the only "attribute" is the user's identity. ABE generalises IBE to arbitrary boolean policies over multiple attributes. Lattice IBE is more mature than lattice ABE because the construction is simpler, see [PQ Identity-Based Encryption](pq-identity-based-encryption).

### How does ABE compare to using public key encryption with N recipients?

For small recipient sets known at encryption time, encrypting once per recipient (or using multi-recipient encryption with a fixed list) is often simpler and faster than ABE. The threshold where ABE wins is roughly when the recipient set is large (50+) and dynamically changing, or when the policy is more expressive than a list. For most enterprise file sharing, multi-recipient encryption is the right tool. For cloud storage where recipients are determined by future attribute assignments, ABE is the conceptual fit.

### Are there any open-source lattice ABE libraries?

Several research codebases exist. The OpenABE library (originally from Voltage, now community-maintained) supports pairing-based ABE but not lattice ABE. Academic prototypes exist at MIT, ENS, and a few other research groups, but none has reached production maturity. Building lattice ABE into a mainstream library like OpenSSL or libsodium would require both standardisation (currently absent) and significant performance work.

## Sources

1. Sahai, A., Waters, B. "Fuzzy Identity-Based Encryption." EUROCRYPT 2005. https://eprint.iacr.org/2004/086
2. Goyal, V., Pandey, O., Sahai, A., Waters, B. "Attribute-Based Encryption for Fine-Grained Access Control of Encrypted Data." CCS 2006. https://eprint.iacr.org/2006/309
3. Gorbunov, S., Vaikuntanathan, V., Wee, H. "Attribute-Based Encryption for Circuits." STOC 2013. https://eprint.iacr.org/2013/337
4. Boneh, D., Gentry, C., Gorbunov, S., et al. "Fully Key-Homomorphic Encryption, Arithmetic Circuit ABE, and Compact Garbled Circuits." EUROCRYPT 2014. https://eprint.iacr.org/2014/356
5. Datta, P., Pal, T., Yamada, S. "Efficient Lattice-Based ABE." CRYPTO 2021. https://eprint.iacr.org/2021/356
6. Agrawal, S., Yamada, S. "Optimal Broadcast Encryption from LWE and Pairings in the Standard Model." TCC 2020. https://eprint.iacr.org/2020/1304
7. Chase, M. "Multi-Authority Attribute-Based Encryption." TCC 2007. https://www.cs.brown.edu/~mchase/papers/multiabe.pdf
8. NIST. "Privacy-Enhancing Cryptography Project." https://csrc.nist.gov/Projects/pec

## Related Articles

- [What Is Post-Quantum Cryptography?](what-is-post-quantum-cryptography)
- [ML-KEM Explained](ml-kem-explained)
- [Lattice-Based Cryptography Explained](lattice-based-cryptography-explained)
- [PQ Identity-Based Encryption](pq-identity-based-encryption)
- [PQ Functional Encryption](pq-functional-encryption)

---

### Protect Your Data Before Q-Day Arrives

QNSQY's NIST-standardized post-quantum encryption protects files against both current and quantum-era threats.

[Try QNSQY](../../pricing)
