Adtech & Antifraud Glossary
Short, plain-language definitions for the terms you’ll run into evaluating click and impression fraud detection.
C
Click Fraud
Illegitimate clicks on an ad or affiliate link generated by bots, click farms, or automated scripts, intended to drain ad budget or fake performance without any real user interest.
click_id
The unique identifier your tracker (Keitaro, Binom, Voluum, etc.) generates for each click, usually passed as a `{clickid}` macro. Antifraud postbacks key off this ID so the verdict lands on the right record.
D
Device Fingerprinting
Identifying a device from a combination of browser and hardware attributes (screen resolution, fonts, user agent, WebGL signature, etc.) without relying on cookies. Used to spot repeat bad actors even when they clear cookies.
F
Fallback URL
The destination a traffic-back redirect sends fraudulent or suspicious traffic to instead of the real offer — commonly a blank page, a lower-value backup offer, or an internal audit page.
Fingerprinting
See Device Fingerprinting.
G
GIVT (General Invalid Traffic)
Invalid traffic from known, declared, non-malicious sources — search engine crawlers, monitoring bots, and other traffic that identifies itself honestly. Filtered out as routine hygiene rather than treated as an attack.
Graceful Degradation
A design pattern where, if the antifraud service is unreachable or times out, traffic is allowed through by default instead of being blocked — so an outage on the vendor side never stops your funnel.
I
IVT (Invalid Traffic)
The umbrella term for all non-human or fraudulent traffic — bots, automation tools, and scammer-generated hits. Splits into GIVT (general, declared) and SIVT (sophisticated, undeclared).
J
JS Tag
A JavaScript snippet placed on a landing page that collects visitor signal (behavior, device, environment) client-side for fraud analysis.
P
Page Acceptance
A behavioral metric indicating whether a visitor spent more than a set threshold (commonly 15 seconds) of active time on a page — used to catch very short, bot-like visits.
Page Activity
A behavioral metric indicating whether any user interaction (mouse movement, scroll, click, or touch) was recorded during a visit. Absence of activity is a strong bot signal.
Page Visibility
A behavioral metric based on the browser’s Page Visibility API, indicating whether a tab was ever brought into focus during a visit.
Ping-Post
A lead-distribution pattern (common in insurance/finance verticals) where a lead is "pinged" to multiple buyers for a bid before being "posted" in full to the winner. Distinct from, but often confused with, postback/ping-back for fraud verdicts.
Postback
An asynchronous server-to-server POST request that reports an event or verdict from one system to another — e.g. an antifraud engine notifying a tracker that a given click_id was scored as fraud. Also called ping-back or S2S callback.
Pull Model
A postback architecture where the client requests the verdict from the antifraud system within a limited time window (TTL), rather than the antifraud system pushing it automatically. Requires the client to run a polling/scheduling job.
Push Model
A postback architecture where the antifraud system sends the verdict to the client’s callback URL as soon as it is available, with no polling required on the client side.
R
Redirect Integration
An antifraud setup where the tracker link points at the antifraud system first; it inspects the request, then 302-redirects clean traffic to the real offer and fraud traffic to a fallback URL. Requires no landing-page code changes.
S
S2S (Server-to-Server)
Any data exchange that happens directly between two backend servers rather than through a user’s browser — postbacks, ping-backs, and most tracker-to-antifraud integrations run S2S.
SIVT (Sophisticated Invalid Traffic)
Invalid traffic that deliberately hides its non-human or fraudulent nature — spoofed user agents, headless browsers, click farms, botnets. Requires active detection rather than a static blocklist.
T
Traffic-Back (Redirect)
A feature that automatically routes fraud-flagged traffic to a fallback URL at the redirect layer, before it ever reaches the real landing page.
TTL (Time to Live)
The time window during which a stored fraud verdict remains available to be queried before it is discarded. Relevant mainly to pull-model postback architectures, where missing the TTL means the verdict is lost.