Why it matters
- Deepfakes of executives can damage reputation and stock prices
- Synthetic media impersonation is growing exponentially
- Brand teams need verifiable evidence packages to act on suspected impersonation
What Raina does
Visual deepfake detection
Frame-level synthetic-image probability from the universal forensic pipeline, surfaced as part of every brand-vertical scan.
Voice clone detection
Per-clip synthetic-audio probability via Resemble Detect and AssemblyAI Voice. The brand panel reports it as a separate signal from visual deepfake probability.
Protected-assets registry
Per-organization registry of executive headshots, logos, voice clips, and registered taglines. Scans match against the registry and surface the matched asset in the evidence package.
Continuous monitoring
Opt-in per asset. The scheduler runs every 15 minutes, learns a rolling mention baseline, and opens an incident when amplification spikes past the threshold.
Incident tracking + webhooks
Each spike opens an incident record with a triggered-snapshot audit. A signed webhook fires to the configured URL so downstream tools can route the alert.
Evidence package builder
Single-click bundle of headline, summary, signal block, and routing recommendations for legal, platform-abuse, and comms teams.
Incident triage and routing
Each amplification spike opens an incident with a webhook delivery record and a status workflow (open → acknowledged → resolved or dismissed). Analysts can route incidents to legal, platform abuse, or comms with a signed evidence package attached.
How it's verifiable
C2PA
Content credentials and provenance verification on submitted media.
Ed25519
Every verification report and evidence package is cryptographically signed.
Bayesian
Source-tier scoring model under the corroboration engine.
Example workflows
Profile setup
Register protected assets, executives, and brand elements
Continuous monitoring
Opt-in per asset. The scheduler watches social amplification and forensic match signals between scans.
Triage and route
Analyst opens the incident in /brand-incidents, reviews the trigger snapshot (current mentions vs baseline, webhook delivery status), and either acknowledges, resolves, or dismisses. Status transitions are audited.
Verification on demand
Submit suspected impersonation content for multi-layer authenticity analysis. Visual deepfake and synthetic-audio probabilities are surfaced as separate signals.
Evidence package
Receive a signed verification report and an evidence package formatted for legal action and platform abuse reporting.