{ "email": "user+promo2@gmail.com", "deliverability": "DELIVERABLE", "quality_score": 0.15, "is_free_email": true, "is_disposable_email": false, "is_subaddress": true, "is_role_email": false, "is_smtp_valid": true }






Fraud detection is the practice of spotting bad activity before it causes loss: fake signups, stolen cards, bot traffic, and abuse of promos or accounts. It works by scoring the identity and context behind each action against known risk signals, then flagging or blocking the ones that look fraudulent.
Detection identifies fraud that is happening or has happened, usually by scoring signals and surfacing risky activity. Prevention stops it before it lands, by blocking or challenging risky actions at signup, login, or checkout. Most teams need both: detection to see the fraud, prevention to act on it.
Common types include fake and new-account fraud, bot attacks, SMS pumping, promo and referral abuse, chargeback and payment fraud, account takeover, ad fraud, and multi-accounting. They differ in target but share the same tells: throwaway emails, VOIP numbers, and VPN or datacenter IPs.
Start with signal checks at the entry points, signup, login, and checkout, to catch the obvious tells cheaply. Layer on rules or an ML platform for scoring and case management as volume grows. Abstract provides the signal layer; it is not a full fraud platform or SIEM, so pair it with those for orchestration and investigation.
Abstract is the data layer, not a full fraud platform. Its Email Validation, Phone Validation, and IP Intelligence APIs return the signals, disposable emails, VOIP numbers, VPN and proxy IPs, that reveal fraud. You combine those signals and decide the action, or feed them into your existing fraud tooling.
The highest-value tells are a disposable or free email, plus-addressing, a VOIP or invalid phone number, and a VPN, proxy, or datacenter IP. No single flag is proof, but several together on one action are a strong fraud signal you can act on with confidence.