Free Ebook

20 Common Technical Fraud Signals ebook cover

20 Common Technical Fraud Signals

A field guide for sample buyers

We break the black box of "technical signals" into five types of fraud signals, then walk through the 20 signals inside them, covering what each detects, where each is ambiguous, and how to tell a vendor that understands its own tooling from one reading off a feature list.

5 types of signals20 signals explained7 questions to ask vendors

Get the free ebook

Fill out the form and the download unlocks instantly.

What's Inside

Five types of signals, twenty signals total

Every signal in the guide is organized by the question it tries to answer. Each type gets its own chapter, with what each signal detects and where it gets ambiguous.

Type 01

Duplication & identity continuity

New respondent, or the same actor in a new form?

  • Duplicate fingerprints
  • Project-level overlap
  • Subnet patterns
  • Prior behavior history

Type 02

Network concealment & routing

Where did this traffic really come from?

  • VPN usage
  • Tor
  • Public proxy
  • Web proxy

Type 03

Automation & synthetic environments

Is software steering the session instead of a person?

  • Emulators
  • AI & agent orchestration
  • navigator.webdriver
  • Developer tools

Type 04

Reputation & historical abuse

What has this environment been tied to before?

  • Crawler history
  • Internet fraud history
  • Retail & ad-tech fraud
  • Recent abuse

Type 05

Orchestration & obfuscation

Is this one attempt, or a coordinated, concealed operation?

  • External risk-score failure
  • WebRTC inconsistencies
  • Batch entrants
  • Obfuscation mechanisms

A signal is a clue, not a verdict, and treated as a verdict almost every one produces false positives. The number of signals tells you little. The way they are combined, weighted, and explained tells you most of what you need to know.

Want signals read the right way, on your studies?

The guide teaches you the questions. Our team runs the answers on billions of survey attempts a year, combining and weighting signals so clues never get treated as verdicts.