← All guides
Proxy & VPN detection

Fraud filtering for affiliate trackers: Keitaro, Binom, RedTrack and Voluum

Score every click before it reaches the offer. What each tracker can do natively, where it stops, and the one-file gate that works with all four.

Published

If you send paid traffic through Keitaro, Binom, RedTrack or Voluum, your tracker decides where each click goes, but not whether the person behind it is real. This guide explains what each of the four trackers can do about proxy, VPN and bot traffic on its own, where each one stops, and how one small self-hosted gate adds a fraud score to every click in all four.

It is written for affiliates and media buyers in Pakistan, India, Bangladesh and everywhere else. You do not need to be a developer, only a small host where you can upload one script.

Why a click needs a score before the offer

By the time a lead reaches your affiliate network, the damage from a bad click is done: you paid for it, and the network scrubbed the conversion because it came from a data center, a VPN exit or a bot. Do that often enough and the network starts to question your whole source.

The cheap place to stop that traffic is before the offer loads. A visitor who never sees the offer cannot become a rejected conversion, and a click you turn away can still be counted in your tracker, so you see which placements send the junk.

A FraudNova check on a visitor's IP returns a fraud_score from 0 to 100 and a set of true/false flags, including proxy, vpn, tor and bot_status. The proxy and VPN detection API page lists every field. The question is how to get that check into your tracker's click path.

What each tracker can and cannot do

We read the official documentation of all four. Each has some built-in protection, none asks an outside scoring service about every click and branches on the answer, and each can send a visitor through a URL you control with the click id attached. That last point is what makes one shared solution possible.

Keitaro runs on your own server and is written in PHP. Its Proxy filter reads the request's HTTP headers (Keitaro says it works better with the IP2Location PX2 database), its Bot filter matches a bot list, and the documentation states plainly that the tracker has no VPN filtering. What sets Keitaro apart is that it loads custom PHP filters, so it is the one tracker where the check can run inside the tracker itself.

Binom is self-hosted too, with Binom Cloud as the hosted option. Its campaign rules include a Proxy-traffic rule, whose data source the documentation does not describe, and a Bot rule that works with MagicChecker, Adspect or Binom Protect. Rules can also route on a value in the URL's GET parameters. There is no filter plugin, and no rule calls an external API. Its Click API lets a PHP lander run code of its own, which you then maintain yourself.

RedTrack is cloud only, so there is nothing to install inside it. Its filters cover proxy types (Anonymous, Transparent and Hosting) and list a "Fraud check" option that the help pages do not explain, and Blacklist bots blocks IP ranges and user agents you supply. Filtering needs redirect tracking, and RedTrack documents no code hook.

Voluum is cloud only as well. Its Anti-Fraud Kit reports suspicious visits rather than stopping them; blocking is done with IP/UA filtering and rule-based paths, including a Proxy rule. Voluum also has a Redirect Webhook, which calls your server when a campaign URL is hit and uses the lander or offer address you send back within 1 second, but Voluum does not publish its request and response format.

The gate, explained once

The gate is a one-file redirect script you host yourself. It lives in the tracker gate folder of our examples repo, in two versions that behave the same: PHP (needs PHP 8.2 and Composer) and Node.js (needs Node.js 18 or newer). The tracker sends traffic to the gate instead of your lander, and the gate forwards each visitor to your money URL or your safe URL.

It reads three things from the visitor's own request: the real connection IP, the User-Agent header and the Accept-Language header. It never takes the IP from a query parameter, so nobody can buy a clean verdict by adding ?ip= to a link. It checks the IP with FraudNova, waiting at most 2 seconds, and caches the verdict per IP for 24 hours.

The click is blocked when the fraud_score is 75 or more, or when proxy, vpn, tor or bot_status is true. You can change the number and the list of flags, add active_vpn, and set allowed_countries so traffic outside the markets you buy is turned away too. Either way the gate answers with a 302 redirect and carries the incoming query string along, so the tracker's click id survives the hop, and any {clickid} placeholder in a destination URL is filled in with it.

If the check fails for any reason (you are out of credits, you hit a rate limit, the service has an upstream problem, or the answer takes longer than 2 seconds), the click is allowed. A good visitor is never sent away because a check did not run, though you can choose to fail closed. And if you set report_url to your tracker's postback, the gate calls it once for every blocked click, so the tracker shows blocked clicks as a custom conversion next to the real ones.

Which tracker needs it? All four can use the gate. On RedTrack it is the only way to score each click, and on Binom the only one that does not mean writing lander code yourself. On Keitaro you can choose between the gate and a custom filter that runs the same rules inside the tracker. On Voluum the gate is the lander route available today; the Redirect Webhook could call a server like the gate once Voluum gives you its request and response format.

Choosing a threshold

FraudNova scores fall into three bands:

  • 0–39, clean. Send it to the offer.
  • 40–74, suspicious. Worth watching on expensive offers.
  • 75–100, high risk. The band you block.

Think of the threshold as a dial between two costs: lower turns away more real visitors, higher lets more junk through. Leave it at the default of 75, the first high-risk score, until your own numbers say otherwise. Offers that pay well but scrub hard (finance, insurance, strict lead generation) can spare a few borderline visitors, so 60 or 65 suits them. On cheap, high-volume traffic a stray suspicious click costs little, so 85 keeps out only the clearest offenders. Either way, leave the proxy and VPN flags on: networks that refuse that traffic ignore its score. The fraud score thresholds guide covers tuning it with your own data.

What it costs

You pay for unique visitors, not clicks. Each check is 1 credit and the gate remembers each verdict for 24 hours, so only the first click from an IP that day is charged. A campaign with 5,000 clicks a day from 3,000 different IPs uses about 3,000 credits. Credits come in prepaid packs with no subscription, a verified email adds 25 free ones, and HEAD requests from link checkers go to the safe page at no cost. Rate limits and the concurrent-check cap are in the API docs.

Set it up in your tracker

Each tracker puts the gate in a different place and has its own click-id token and postback format. Pick yours:

  • Keitaro fraud filter: the custom PHP filter, or the gate as a stream's landing page, with blocked clicks reported as trash.
  • Binom fraud filter: the gate as the campaign lander with "Append to lander's URL", and how it works alongside the Proxy-traffic rule.
  • RedTrack fraud filter: the gate as a lander on a redirect-tracking campaign, with a blocked conversion type.
  • Voluum fraud filter: the gate in a Lander → Offer path, a blocked custom conversion, and what we know about the Redirect Webhook.

Next step

Start with an address you already doubt: paste it into the free IP checker and see the score and flags the gate would act on. Then create an account for an API key, set the gate up with your tracker's guide, and route one campaign through it while the rest run unchanged. After a few days, set its blocked clicks and conversion rate against the others.

Try it free

Run one IP proxy check right now — no account, no card.

Try the free IP proxy checker — no signup See the Proxy & VPN detection API

Also read