Machine translation, not reviewed by a native speaker. If in doubt, the French text prevails.

Security

Dashvia security, explained plainly

In one sentence: Dashvia reads your data without ever changing it, has no accounts of its own (identity comes from Idenva), applies row-level rights to every query and records who saw what in a log the database itself refuses to rewrite.

1. Read-only sources

SQL sources are opened read-only. A write is refused by the engine, not only by the interface; the safeguard is proven with a passing case, a violating case and a deliberate sabotage.

2. Federated identity, no local accounts

Sign-in goes through Idenva (SSO). A token from another issuer, or an expired one, is refused, and re-enabling a local password is detected. Dashvia keeps no user database of its own.

3. Row-level rights

Idenva groups decide which rows are visible. The filter is parameterised, never built by interpolating values, and it also applies to cross-filters, drill-down, exports and scheduled reports.

4. An AI that shows its query

Text-to-SQL answers with a readable SQL query. Any read outside the schema and any write is refused before execution. Without the Kopilvia hub, Dashvia switches to an explicit degraded mode: local corpus, and a refusal to answer outside it rather than a guess.

5. Audit log

Who viewed, exported or changed which report is recorded in an append-only log: the database refuses any update or deletion of a row already written.

6. Hardened, pinned base

Dashvia builds on Apache Superset (Apache-2.0), pinned at version 6.1.0. A default secret key or an admin/admin account is refused at start-up, and published security advisories are checked against the pinned version.

7. Self-hostable

The base installs on your own infrastructure. Your schemas and results do not have to leave it.

A question?

Write to us: we also answer technical questions.