Impression Fraud

Impression fraud detection

Every impression is scored for invalid traffic using IAB categories — GIVT and SIVT. Fraud is caught, shown by reason and source, and turned into protection.

Two categories (IAB)

GIVT General Invalid Traffic
Deterministic, high-confidence signals decided instantly: datacenter/hosting IPs, known bots and crawlers, headless browsers, addresses already in the block-list.

SIVT Sophisticated Invalid Traffic
Decided by combined signals and cross-visit patterns: proxy/VPN, geo mismatch, non-viewable fast bounce, impression flooding from one IP or subnet, hidden iframes.

How scoring works

Cheap deterministic GIVT checks run synchronously the moment an impression arrives — an instant verdict with no extra lookups. Heavier SIVT analysis (velocity across IPs and subnets, cross-visit patterns) runs asynchronously in a background worker. Each impression ends up as clean, GIVT, or SIVT with the exact reasons attached.

Impression→ GIVT sync→ SIVT worker→ verdict + reasons

The "Impressions" dashboard

Everything lands in one section of your account:

Rates — total impressions, viewability rate, invalid (IVT) rate split into GIVT and SIVT.

Reasons — which signals fired, tagged GIVT or SIVT, with counts.

Geography & devices — invalid share by country, and by platform / mode (creative vs publisher).

Top publishers by fraud — the domains where your budget leaks, sorted by invalid share. Turn off the dirty ones.

From detection to blocking

Fraudulent IPs are automatically added to your block-list. A sync worker uploads the block-list to Yandex.Direct and Google Ads as IP exclusions — so those addresses stop receiving your ads. This blocks future impressions from known-fraud IPs.

Fraud detected→ block-list→ Yandex / Google→ no more ads to that IP

Note: the pixel measures — the current impression has already happened. Blocking applies to future impressions. Real-time blocking before the bid is available via the Pre-Bid API.

Getting started

Install the impression pixel, and detection starts immediately — or reach us at support@mediasafe.cc.