Encryption model & independent audits
How this is scored
What the vendor can prove about its security design: end-to-end encryption with named algorithms and key derivation, which fields are encrypted on the device and which are not, independent audits and penetration tests with published results, a bug bounty, and incident history disclosed.
0 — Vault contents are readable by the vendor by its own description: server-side encryption with keys the vendor holds, or no encryption of stored credentials.
3 — Client-side or end-to-end encryption described with the algorithms named; key derivation, audits and the scope of what is encrypted are basic or not described.
5 — A published security whitepaper or security page naming algorithms, key derivation and the client-side encryption model, and an independent audit, penetration test or certification (ISO 27001, SOC 2) named.
8 — A stated list of which fields are and are not encrypted (including URLs and metadata), recurring independent penetration tests or code audits with dates and reports or summaries published, a public bug bounty or vulnerability disclosure policy, past security incidents documented with their impact, and open-source clients or cryptography that can be reviewed.
10 — The design is verifiable end to end: full source or cryptographic design public, every field encrypted client-side including metadata, audits by named firms repeated yearly with full reports, a threat model that states what a compromise of the vendor's servers would expose, and incident post-mortems published.
The DevOps Engineer
The security page names the algorithms (AES-256, RSA-2048, elliptic curve), the master-password key derivation (PBKDF2 with one million iterations) and which record contents are encrypted down to URLs, attachments and TOTP codes, backed by FIPS 140-3 validated modules, named external testers (NCC Group, CyberTest) engaged periodically, a Bugcrowd-run bounty and annual accredited ISO 27001 audits. The captured pages give me no dated penetration-test reports, no incident history and no open-source client code I could read before trusting the cryptography, and as someone who audits before deploying, that caps the score. 5 6 12 5