BridgeKeeper

A passwordless authentication system that replaces traditional passwords with personal memories. Users log in by answering challenge questions spoken aloud — the system verifies both the semantic meaning of the answer and the speaker's voiceprint, collapsing two authentication factors into a single natural interaction.

How it works

• During onboarding, users create 10 personal challenge/response pairs
• At login, the system randomly selects 3 challenges (720 possible sequences)
• Answers can be spoken (dual-factor: semantics + voice biometrics) or typed (semantic only)
• Semantic similarity is computed via AI embeddings, not exact string matching
• Voiceprint verification confirms the speaker's identity through vocal characteristics
• Magic link email fallback when voice authentication is unavailable
• Individual challenge pairs can be rotated if compromised

Security model

• No passwords are stored anywhere in the system
• Voice embeddings are stored as hashed vectors, not raw audio
• All biometric data is encrypted at rest and in transit
• Rate-limited to 3 failed attempts per hour
• Trust levels vary by method: high (voice), medium (text), medium (magic link)
• Designed with GDPR Article 9 biometric data compliance in mind

Tech stack