Searchable Encryption with Postgres Types
Presented by:
Toby Hede
CipherStash
No video of the event yet, sorry!
Encrypt Query Language (EQL) brings searchable encryption to PostgreSQL as ordinary column types. Encrypted values stay encrypted in the database, yet still support equality, ordering, and range queries backed by real indexes — with no decryption. This talk shows how encrypted types, index terms, and functional indexes fit together, and how a single source of truth keeps their query behaviour consistent across every supported type.
What attendees will learn (about Postgres):
- Domains as a design tool — using domain types and check constraints to model a column that carries more than a plain value, and the limits you hit (e.g. why you can't stack a domain on a domain).
- How operators actually resolve — why = and < on a custom type behave the way they do, and how to control it — including making an unsupported operation fail loudly instead of silently returning a wrong answer.
- Functional indexes and the planner — how an ordinary functional index makes bare WHERE col = $1 engage without operator classes or superuser, so it works on managed Postgres.
- Keeping a large type surface honest — treating a family of related types as a single source of truth, and testing their query behaviour against real data rather than assumptions.
- The takeaway — a reusable mental model for pushing Postgres's type system, using searchable encryption as the stress test.
- Date:
- Duration:
- 40 min
- Room:
- Conference:
- PG Down Under 2026
- Language:
- Track:
- Development
- Difficulty:
- Medium