GuidesIntelligenceFraud Detection (Defender)

Fraud Detection (Defender)

Defender runs 20 fraud checks on every bid request at 0ms added latency. All checks execute in-process at the edge — no external calls, no performance impact.

How It Works

Every incoming bid request passes through Defender before the auction runs. Invalid traffic is blocked immediately. Legitimate requests proceed to auction as normal.

Bid Request → Defender (20 checks, 0ms) → Auction
                ↓ (if blocked)
              204 No Bid

Fraud Checks

Device Fingerprinting (16 signals)

Defender analyzes 16 device signals to detect spoofed or fabricated devices: user agent consistency, screen resolution, language, timezone, connection type, device model, OS version, and more. Mismatches between signals indicate bot or spoofed traffic.

Network Checks

CheckWhat It Catches
Data center IPTraffic from hosting providers, not real users
Tor exit nodesTraffic routed through Tor network
VPN detectionKnown VPN provider IP ranges
Proxy detectionOpen proxy and residential proxy networks

Supply Chain Verification

CheckWhat It Catches
SSAI spoofingFake server-side ad insertion headers
Domain spoofingMismatched domains in bid request vs actual site
ads.txtPublisher not authorized to sell this inventory
Supply chain (schain)Broken or missing seller chain

Behavioral Analysis

CheckWhat It Catches
Geo anomaliesIP location contradicts declared geo
Time anomaliesTimezone mismatch with declared location
Pacing anomaliesUnnaturally even request distribution
Bot detectionKnown bot signatures and crawlers

Auto-Blacklisting

IPs that repeatedly fail fraud checks are automatically blacklisted. Blacklisted IPs receive an immediate 204 response with no auction.

Dashboard

Go to Intelligence > Defender in the sidebar.

KPIs

  • Requests scanned — total requests checked today
  • Blocked — requests rejected by Defender
  • Block rate — percentage of traffic blocked
  • Auto-blacklisted IPs — IPs added to blacklist today

Hourly Trend

A chart showing blocked vs allowed traffic by hour. Use this to spot sudden spikes in fraudulent traffic.

Type Breakdown

A breakdown of blocks by check type (e.g., 40% data center IP, 25% bot detection, 15% domain spoofing). Helps identify the main sources of invalid traffic.

Via Broker Chat

“What’s the fraud rate today?”

“Show me Defender stats for the last 7 days”

“Which publishers have the highest block rate?”