Search over encrypted data using fully homomorphic encryption. Build encrypted indexes, run keyword and similarity queries on ciphertext. Zero plaintext exposure at any point in the pipeline.
Documents and fields are encrypted with BFV or CKKS before indexing. The encrypted index is built server-side over ciphertext. Original data never appears in plaintext on the search infrastructure.
Your search query is encrypted client-side and sent to the server. FHE operations match the encrypted query against the encrypted index. The server computes on ciphertext without decrypting anything.
Search results are returned as encrypted ciphertexts. Only the data owner can decrypt the matching records. The search server never learns what was queried or what matched.
curl -X POST https://api.h33.ai/api/search/index \ -H "Authorization: Bearer $H33_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "index_name": "patient_records", "documents": [ { "id": "doc_001", "fields": { "name": "encrypted_bfv_ct...", "diagnosis": "encrypted_bfv_ct...", "dob": "encrypted_bfv_ct..." } } ], "schema": { "name": "keyword", "diagnosis": "keyword", "dob": "range" } }'
{
"index": "patient_records",
"documents_indexed": 1,
"fields_encrypted": 3,
"index_size_bytes": 4096
}
curl -X POST https://api.h33.ai/api/search/encrypted \ -H "Authorization: Bearer $H33_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "index_name": "patient_records", "query": { "type": "keyword", "field": "diagnosis", "value": "encrypted_query_ct..." }, "limit": 10 }'
{
"hits": [
{
"id": "doc_001",
"score": 1.0,
"fields": {
"name": "encrypted_ct...",
"diagnosis": "encrypted_ct..."
}
}
],
"total": 1,
"search_latency_ms": 2.4,
"decryption_required": true
}
curl -X GET "https://api.h33.ai/api/search/query?index=patient_records" \ -H "Authorization: Bearer $H33_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "query": { "bool": { "must": [ { "keyword": { "field": "diagnosis", "value": "encrypted_ct_diabetes..." } }, { "range": { "field": "dob", "gte": "encrypted_ct_1980...", "lte": "encrypted_ct_2000..." } } ] } } }'
{
"hits": [...],
"total": 3,
"search_latency_ms": 4.1,
"boolean_ops": 2
}
FHE-powered exact and partial keyword matching over encrypted fields. BFV integer encoding enables precise string comparisons without decryption. Supports prefix, suffix, and substring patterns.
Run cosine similarity and nearest-neighbor searches on encrypted vector embeddings using CKKS approximate arithmetic. Ideal for semantic search over encrypted documents and records.
Combine encrypted search predicates with AND, OR, and NOT operators. Build complex multi-field queries that execute entirely in ciphertext space. Full boolean algebra on encrypted data.
Build and maintain search indexes over encrypted data. Inverted indexes, B-trees, and hash maps all operate in ciphertext. Index updates are encrypted and never expose document contents.
Data is encrypted before it reaches the search server and stays encrypted throughout indexing, querying, and result retrieval. Not even the infrastructure operator can access plaintext.
Meets the technical safeguard requirements for HIPAA, GDPR, CCPA, and SOX. Search sensitive data without creating compliance exposure. Audit-ready encrypted processing logs.
Search patient records, diagnoses, and treatment histories without exposing protected health information. HIPAA-compliant queries over encrypted EHR data. Researchers can query datasets they never see in plaintext.
Run e-discovery queries across encrypted document repositories. Search privileged communications without exposing attorney-client content. Keyword and pattern matching on sealed court records.
Search encrypted transaction logs for fraud detection, compliance checks, and audit trails. Query encrypted account records without exposing customer financial data to search infrastructure.
Search classified and compartmented datasets without exposing content to search operators. Encrypted queries on encrypted intelligence databases. Need-to-know enforcement at the cryptographic level.
| Capability | Traditional Search | H33 Encrypted Search |
|---|---|---|
| Data at rest | Decrypted for indexing | Always encrypted |
| Data during search | Plaintext in memory | Ciphertext throughout |
| Query privacy | Server sees queries | Queries are encrypted |
| Result privacy | Server sees results | Results are encrypted |
| Index exposure | Reveals data structure | Index is encrypted |
| Breach exposure | Full plaintext leak | Ciphertext only |
| Compliance risk | High (plaintext processing) | Minimal (no exposure) |
| Insider threat | Operators see data | Operators see ciphertext |
All units fungible across H33 products. Free tier includes 1,000 units. No credit card required.
/api/search/query endpoint supports cross-index queries. Results from multiple indexes are merged server-side while still encrypted. Only the client can decrypt the combined results./api/search/index endpoint. Submit documents with pre-encrypted fields and a schema defining the search type per field (keyword, range, similarity). The server builds an encrypted inverted index over the ciphertext.Get your API key in seconds. Build encrypted search into your application on your free tier.
Start Searching Encrypted1,000 free units/month · BFV + CKKS · Zero plaintext exposure