Face Verification That Catches Proxy Test-Takers
InsightFace with ArcFace embeddings confirms the same person is present from enrollment to submission. Periodic snapshots, continuity scoring, and transparent audit trails.
Embedding model
False positive rate
Compliant biometric storage
To the candidate
Proof that the candidate is the candidate
State-of-the-art biometrics, zero browser takeover.
Enrollment snapshot
Capture a reference face at session start — compared against every subsequent snapshot throughout the assessment.
ArcFace embeddings
State-of-the-art face embedding model from InsightFace. Robust to lighting, angle, and hair changes.
Periodic continuity checks
Webcam snapshots at configurable intervals — each compared to the enrollment reference for continuity scoring.
Real-time alerts
Mid-session mismatches trigger interviewer alerts immediately. Review with replay.
Per-signal authenticity score
Face continuity is one of ten integrity signals in the unified authenticity score.
Hashed biometric storage
Face embeddings stored as vectors, not raw images. GDPR/BIPA-aligned retention policies.
Audit trail
Every snapshot timestamped with confidence score — exportable for compliance review.
Browser-native
Works via the Web MediaDevices API — no extensions, no installs, no plugins.
Consent-first flow
Candidates see a clear consent screen before enrollment. Opt-out is a supported org policy.
Built for how modern teams hire
Why ArcFace, not a toy face API
Consumer face APIs are trained on social media faces and struggle with real-world conditions. ArcFace, trained on massive research datasets, handles lighting, angle, and occlusion robustly — the same model used in airport biometrics.
- InsightFace + ArcFace embedding pipeline
- Robust to partial occlusion (glasses, masks off → on)
- < 1% false positive rate in our tests
Invisible monitoring, clear consent
Candidates are told before enrollment that face continuity is active — compliance requires it. After that, capture runs silently without extensions or disruption.
- Clear consent screen pre-enrollment
- No visible indicator during the session
- Configurable capture interval
Biometric storage that passes legal review
Raw images are discarded after embedding extraction by default. Embeddings are stored as vectors with configurable retention, hashed for storage, and supported in deletion workflows.
- Embeddings stored, not raw images (configurable)
- GDPR Article 9 and BIPA alignment
- Right-to-be-forgotten workflow for biometric data
How it works
Candidate consents
Clear screen explains face continuity is active. Candidate agrees to proceed.
Enrollment snapshot
One reference snapshot captured at session start, embedding extracted.
Continuity checks throughout
Periodic snapshots compared to the enrollment embedding — live continuity score.
Review flagged events
Mismatch events surface in the integrity report with timestamp and confidence score.
Frequently asked questions
What model powers face verification?+
InsightFace with ArcFace embeddings — a research-grade face recognition pipeline. ArcFace is the same embedding approach used in airport biometric systems, not a consumer-grade social media face API.
Are raw face images stored?+
By default no — only face embedding vectors are stored. Raw images are discarded after embedding extraction. This is configurable per org for compliance requirements.
What if a candidate adjusts lighting or wears glasses mid-session?+
ArcFace is robust to lighting changes, angle shifts, and partial occlusion (glasses, minor facial changes). Tuning parameters allow orgs to set their own false-positive sensitivity.
Is face verification GDPR and BIPA compliant?+
ClarityHire supports GDPR Article 9 (special category data) and BIPA requirements via clear consent flows, configurable retention, biometric deletion workflows, and audit logging. Check with your legal team for jurisdiction-specific advice.
Can candidates opt out of face verification?+
Yes — opt-out is a configurable org policy. If required by law, candidates can decline biometric capture and your team can choose to proceed with other integrity signals.
Explore related features
ClarityHire is one platform. Every feature is built to work with the rest.
Cheat Detection
Ten signal types — face, keystroke, gaze, code coherence, paste events — analyzed in real time.
Keystroke Biometrics
XGBoost classifier detects when a different person takes over the keyboard.
Code Coherence AI
Claude-powered analysis judges whether code edits reflect authentic human thinking.
Online Proctoring
Browser-native remote proctoring with fullscreen lock, tab tracking, and webcam checks.
Coding Assessments
Live collaborative coding with Monaco, integrated execution, and AI-assisted grading.
Verify your remote candidates are real
Enable face continuity in one click. Research-grade biometrics, zero candidate friction.