Deepfake & impersonation scan

Submit a subject for deepfake and impersonation analysis. Returns scores, verdict, and indicators. Automatically ingests a risk signal if scores exceed the auto-trigger threshold (60).

How detection works

Scans use a heuristic signal engine (heuristic_v1) that accepts caller-provided signals (from external ML models, rule engines, or manual review) and maps them to weighted scores:

• Deepfake indicators: face swap, inconsistent lighting, compression artifacts, temporal inconsistency, metadata mismatch
• Impersonation indicators: biometric mismatch, name mismatch, identity doc mismatch, issuer trust violation

Scores ≥ 75 set verdict deepfake or impersonation. Scores ≥ 40 set suspect. Scores ≥ 60 automatically ingest a risk signal.

You can also bypass heuristic analysis by passing pre-computed deepfake_score and impersonation_score values directly. The platform still applies the same verdict thresholds and automatic signal ingestion.

See the deepfake detection guide for the full workflow, indicator reference, and additional examples.

Request

subject_ref · string · required

Reference to the subject being scanned — URL, SHA-256 hash, attestation ID, or any opaque identifier.

subject_type · string · required

Type of subject: image, video, document, attestation, identity

signals · array

Array of indicator strings from your external model or rules engine (e.g. ["face_swap_detected", "metadata_mismatch"]). When provided without pre-computed scores, the heuristic engine maps these to weighted deepfake and impersonation scores.

attestation_id · string

UUID of an attestation to associate with this scan result. Use this to tie detection results to the credential they relate to.

deepfake_score · integer

Pre-computed deepfake score (0–100) from an external model. If provided along with impersonation_score, heuristic analysis is skipped.

impersonation_score · integer

Pre-computed impersonation score (0–100) from an external model.

Response

scan_id · string

UUID of the scan result.

subject_ref · string

The subject reference that was scanned.

subject_type · string

Type of subject: image, video, document, attestation, identity

deepfake_score · integer

Deepfake probability score (0–100).

impersonation_score · integer

Impersonation probability score (0–100).

verdict · string

authentic | suspect | deepfake | impersonation

indicators · array

List of triggered indicators from the scan.

detection_model · string

Detection model used (e.g. heuristic_v1).

signal_id · string

UUID of the auto-ingested risk signal (only present when score ≥ 60).

attestation_id · string

UUID of the linked attestation, if one was provided during the scan.

created_at · string

ISO 8601 timestamp of when the scan was performed.