Select a device intelligence provider (e.g., SEON, Sardine, Alloy, or the device intelligence module within your existing IDV vendor stack) and obtain API credentials; most providers offer a JavaScript snippet or mobile SDK for signal collection.
Embed the device intelligence collection script or SDK at the start of the onboarding session, before the identity capture step, to establish a device fingerprint and behavioral baseline.
Submit the session token or device intelligence payload to the provider's risk scoring API alongside the user's submitted identity data; receive a risk score and signal breakdown covering device type, browser environment, IP reputation, VPN/proxy/Tor detection, and emulator flags.
Integrate the device intelligence score as a pre-screen gate: route high-risk device sessions (elevated fraud score, impossible travel, emulator detected) to enhanced scrutiny or denial before consuming IDV transaction credits.
Log the device intelligence signals and scores in the identity proofing audit record alongside the document and biometric verification outputs to support fraud investigations.
Tune decision thresholds using historical fraud data; out-of-the-box risk score thresholds from vendors are starting points and typically require calibration against your user population.
Known gotchas
Device intelligence signals are probabilistic, not deterministic; a high-risk score indicates elevated fraud probability but not certainty — using it as a hard block without a manual review path will create false positives for legitimate users on privacy-protective tools.
Signal collection via JavaScript snippets is blocked by aggressive ad blockers and browser privacy modes; plan for a meaningful percentage of sessions with degraded signal quality and ensure your risk model handles sparse feature vectors gracefully.
Device fingerprinting raises GDPR, CCPA, and similar privacy regulation questions; confirm with legal counsel whether device signals constitute personal data under applicable law and update your privacy notice accordingly.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp