Building a fraud detection pipeline

A single suspicious signal rarely tells the full story. An email might look fine on its own, but paired with a datacenter IP and a freshly registered domain, the picture changes fast. Combining multiple Veille APIs into a pipeline gives you a composite risk view of every signup.

Architecture overview

At its simplest: validate the email (/v1/intelligence/email), check the domain age (/v1/intelligence/domain), and score the IP (/v1/intelligence/ip). Three calls, three signals, one composite score.

Quick integration

Python

import requests

API_KEY = "YOUR_API_KEY"
BASE_URL = "https://api.veille.io/v1"


def score_signup(email: str, ip: str) -> dict:
    headers = {"x-api-key": API_KEY}

    email_data = requests.get(
        f"{BASE_URL}/intelligence/email", params={"query": email}, headers=headers
    ).json()

    domain_data = requests.get(
        f"{BASE_URL}/intelligence/domain", params={"query": email_data["domain"]}, headers=headers
    ).json()

    ip_data = requests.get(
        f"{BASE_URL}/intelligence/ip", params={"query": ip}, headers=headers
    ).json()

    age = min(domain_data.get("domain_age_days", 365), 365)
    domain_risk = ((365 - age) / 365) * 100

    composite = (
        email_data["risk_score"] * 0.40
        + domain_risk * 0.35
        + ip_data["threat_score"] * 0.25
    )

    return {
        "composite_score": round(composite),
        "disposable_email": email_data["disposable"],
        "vpn_or_proxy": ip_data["is_vpn"] or ip_data["is_proxy"],
        "domain_age_days": domain_data.get("domain_age_days"),
    }

TypeScript

const API_KEY = "YOUR_API_KEY";
const BASE_URL = "https://api.veille.io/v1";
const headers = { "x-api-key": API_KEY };

async function scoreSignup(email: string, ip: string) {
  const [emailResponse, ipResponse] = await Promise.all([
    fetch(`${BASE_URL}/intelligence/email?query=${email}`, { headers }),
    fetch(`${BASE_URL}/intelligence/ip?query=${ip}`, { headers }),
  ]);
  const [emailData, ipData] = await Promise.all([
    emailResponse.json(),
    ipResponse.json(),
  ]);

  const domainResponse = await fetch(
    `${BASE_URL}/intelligence/domain?query=${emailData.domain}`,
    { headers }
  );
  const domainData = await domainResponse.json();

  const age = Math.min(domainData.domain_age_days ?? 365, 365);
  const domainRisk = ((365 - age) / 365) * 100;

  return {
    compositeScore: Math.round(
      emailData.risk_score * 0.4 + domainRisk * 0.35 + ipData.threat_score * 0.25
    ),
    disposableEmail: emailData.disposable,
    vpnOrProxy: ipData.is_vpn || ipData.is_proxy,
    domainAgeDays: domainData.domain_age_days,
  };
}

Recommended weights

Signal	Weight	Why
Email risk score	40%	Disposable/role accounts are the strongest fraud indicator
Domain age risk	35%	Newly registered domains correlate heavily with fraud
IP threat score	25%	VPNs are also used by legitimate privacy-conscious users

5 use cases

1. User registration - compute a composite score at signup and auto-reject above a threshold.
2. Checkout verification - run the pipeline before processing a payment to catch synthetic identities.
3. Marketplace seller onboarding - screen new merchants by combining email, domain, and IP signals.
4. Content moderation - prioritize review for posts from high-risk accounts.
5. Webhook security - validate the origin IP and domain of incoming webhooks before processing them.