Invalid traffic & fraud · also called IVT detection, ad fraud detection
Fraud detection
Fraud detection in advertising is the process and technology of identifying invalid or fraudulent impressions, clicks, installs and conversions, so they can be blocked in advance or excluded from billing.
Fraud detection answers one question at enormous speed: is this ad opportunity, click or conversion real? It combines data and rules to sort genuine human activity from bots, spoofing and manipulation, and it happens at two moments: before buying, via pre-bid filtering, and after serving, via post-bid measurement.
The work follows the GIVT / SIVT split. GIVT filtration uses lists and parameters: the IAB/ABC Spiders & Bots List, data-center IP ranges and invalid user agents. SIVT detection needs multi-point corroboration: IP and proxy reputation, device and browser device fingerprinting, headless browser and automation traits, behavioral analysis of interaction, supply-chain checks with ads.txt and sellers.json, click-to-install time distributions and device-ID resets in mobile, and anomaly detection across whole traffic sources. Vendors also use honeypots and threat research to understand new schemes.
Good detection is measured on both sides: catching fraud, and not blocking real people (false positives). The MRC accredits vendors whose methods and processes pass audit, which is why buyers ask for accreditation scope rather than relying on marketing claims.
Think of it like this
Fraud detection is like airport security: quick document checks for everyone, then deeper screening when several small signals together do not add up.
An example
A DSP evaluates 2 million bid requests per second, dropping those from known data-center IPs and unauthorised sellers pre-bid, then a post-bid vendor analyses the impressions it won and flags a 3% SIVT rate for refunds.
Related terms
Pre-bid filtering
Pre-bid filtering is checking each ad opportunity for invalid traffic, fraud or brand-safety risk before a buyer bids on it, so bad impressions are avoided rather than refunded later.
Post-bid measurement
Post-bid measurement is analysing ads after they have been bought and served, checking for invalid traffic, viewability and brand safety in the real environment, to report quality and support refunds.
GIVT (general invalid traffic)
GIVT (general invalid traffic) is invalid traffic that can be identified with routine, list-based or parameter-based filtration, such as known crawlers, data-center traffic and non-browser user agents.
SIVT (sophisticated invalid traffic)
SIVT (sophisticated invalid traffic) is invalid traffic designed to look human or legitimate, which can only be found through advanced analytics, multi-point corroboration and often human review.
MRC IVT guidelines
The MRC IVT guidelines are the Media Rating Council's standards for detecting and filtering invalid traffic in ad measurement, defining the GIVT and SIVT categories that accredited measurers must handle.
Behavioral analysis
Behavioral analysis in fraud detection is examining how a visitor interacts, such as mouse movements, taps, scrolling, typing rhythm and session timing, to tell real humans from bots and scripted activity.
Sources: MRC Invalid Traffic Detection and Filtration Guidelines Addendum (2020 update), TAG Certified Against Fraud program