Also known as: Anti-Fraud Engine, Fraud Prevention System, Risk Engine
A Fraud Detection System (FDS) is the software a broker uses to analyse users, devices, payments, and trading behaviour in real time and automatically flag or block deceptive activity. It is the gatekeeper that decides which sign-ups are genuine and which affiliate conversions get paid.
An FDS ingests dozens of signals per account — device fingerprint, IP and geolocation, KYC and document checks, deposit method, login timing, and trading patterns such as arbitrage or bonus-only play. It combines rules ('block if 5 accounts share one device') with machine-learning models that score each account's fraud probability, then routes high-risk cases to manual review and clears low-risk ones automatically.
For partners the FDS is where commissions live or die. Every lead an IB sends is scored; consistently low-quality, duplicate, or bonus-abusing traffic drags down the partner's fraud metrics and can downgrade their tier or trigger clawbacks. High-quality traffic that funds and trades normally sails through and pays out fast.
Example: an affiliate's new referral registers and qualifies for a $250 CPA. The FDS runs its checks and finds the device fingerprint matches an account banned last month for chargeback fraud. It halts the payout, links the accounts, and opens a case — so the affiliate is not paid on a client the system considers the same fraudulent user returning.
The system collects signals at registration, KYC, deposit, and trade time, then evaluates them with two layers. A rules layer applies hard thresholds — velocity limits, blacklists, device-to-account ratios, mismatched geolocation — that block obvious abuse instantly. A machine-learning layer scores subtler patterns by comparing the account against historical fraud and known-good behaviour.
Each account gets a risk score. Low scores are approved automatically, mid scores may require extra verification, and high scores are held for a human analyst. Because affiliate conversions are tied to accounts, an account's fraud outcome directly gates whether the IB's CPA or revenue share is confirmed, reversed, or paid. Analysts feed decisions back into the models, so the FDS keeps adapting to new abuse tactics.
The FDS gathers device fingerprint, IP/geo, KYC, payment, and behavioural data across the account lifecycle.
Hard thresholds and blacklists block obvious abuse — velocity, duplicate devices, geo mismatches — instantly.
Machine-learning models score subtler patterns against historical fraud and legitimate behaviour.
Low risk auto-approves, medium triggers extra verification, high goes to a human analyst.
The outcome confirms, reverses, or pays the affiliate's commission, and analyst decisions retrain the models.
Why it matters for partnership: The FDS scores every lead you send and gatekeeps your payouts, so duplicate or bonus-abusing traffic downgrades your tier and triggers clawbacks. Clean, genuinely-trading referrals clear fast and protect your standing.
An affiliate sends a lead to Pepperstone that qualifies for a $200 CPA. The FDS runs its checks and finds the new client's device fingerprint matches an account banned last month for chargeback fraud. It halts the payout, links the two accounts, and opens a review case — so the affiliate is not paid on what the system reads as the same fraudulent user returning.
| Approach | Strength | Weakness |
|---|---|---|
| Rules-based | Fast, transparent, easy to audit | Rigid; misses novel patterns |
| ML-based | Catches subtle and new fraud | Opaque; needs data and tuning |
| Hybrid (typical) | Combines both layers | More complex to operate |
Ask your affiliate manager how your traffic is scoring in their fraud system, then optimise your sources toward the segments that fund and trade cleanly.
Assuming manual tricks — cleared cookies, VPNs, fresh emails — can bypass a modern AI-driven engine; the linked signals still surface and void your commissions.
Typically device fingerprint, IP and geolocation, KYC documents, payment method, login timing, and trading behaviour such as arbitrage or bonus-only play — combined into one risk score.
Initial approval can be provisional. If the FDS later links the account to fraud — a chargeback, a matched banned device — it reverses the conversion during its review window.
Not the raw scores, but affiliate managers can share quality feedback. Ask about approval rates and flagged sources so you can improve your lead quality.
False positives happen — a shared IP or a VPN can raise a flag. Genuine clients usually clear after extra verification, which is why clean KYC matters.
No. KYC verifies identity at onboarding; the FDS is broader and continuous, scoring behaviour and linkage across the whole account lifecycle.
Send real, self-motivated traders, avoid incentivised or bonus-hunting sources, never self-register, and keep one device and network per client.
Most IB agreements let a broker withhold or claw back commissions on accounts it flags as fraudulent or abusive. Here's what triggers it, how to read …