Homomorphic Encryption

Homomorphic Encryption (HE) enables computation on encrypted data. enVector leverages HE so the server can compute similarity without decrypting vectors.

What It Enables

• Compute Inner Product (IP) over encrypted vectors.

• Keep data encrypted at rest, in transit, and during use.

• Allow zero-trust processing: the server operates without plaintext access.

Practical Considerations

• Performance: HE is computationally heavier than plaintext. Use batching and dimension planning for best results.

• Precision: Real-number HE schemes support approximate arithmetic suitable for vector similarity.

• Keys: Clients manage secret keys; servers use shared public/evaluation keys for computation.

When To Use Encrypted Queries

• Use ciphertext queries when query privacy is critical (e.g., PII or proprietary prompts).

• Plaintext queries are acceptable if the query content is non-sensitive and you want lower latency.