Detects automated scripts, headless browsers, and bot frameworks at the entry handoff — before a single survey question is answered. Combines browser fingerprinting, interaction timing, and automation signature analysis.
Headless browser and automation framework detection
Impossible timing signals and interaction velocity
JavaScript environment anomalies and WebDriver flags
Catches
AI bots, Selenium scripts, headless Chrome farms, automated survey completion tools
Builds a persistent hardware fingerprint for each respondent using device attributes, browser characteristics, and canvas/audio rendering signatures. Detects shared, cloned, or emulated devices across multiple entries.
Canvas and WebGL rendering fingerprints
Hardware configuration and screen attribute matching
Virtual machine and device emulator detection
Catches
Click farm shared devices, cookie-cleared repeat entries, emulated devices, virtualised environments
Identifies VPN providers, datacenter ASNs, residential proxy services, and Tor exit nodes by analysing IP routing characteristics, ASN ownership, and known anonymisation infrastructure.
VPN provider and datacenter ASN identification
Residential proxy service detection
Tor exit node and anonymisation layer flagging
Catches
VPN masking, geographic spoofing, proxy-based identity cycling, anonymised repeat access
Analyses routing behaviour at the network layer — ASN hop sequences, carrier type, latency patterns, and BGP path anomalies — independent of the IP address itself. Catches botnet-routed traffic that passes standard IP reputation checks.
ASN routing path and carrier signal analysis
Latency and BGP anomaly detection
Botnet infrastructure routing patterns
Catches
Botnet geo-spoofing via residential devices, malware-infected device routing, clean-IP fraud farms
Validates declared geography against network path signals, GPS data (where available), timezone consistency, and browser locale settings. All three sources must agree for a geography claim to be accepted.
Declared geography vs. network path triangulation
Timezone, locale, and language consistency checks
GPS and browser geolocation cross-validation
Catches
Country-of-origin fraud, geographic quota cheating, fabricated demographic geography claims
Maintains a cross-project identity graph that traces respondent tokens, device fingerprints, and network signatures across all deployments within the same project. Catches the same respondent re-entering under different profiles across multiple vendors simultaneously.
Cross-deployment token and fingerprint matching
Cross-vendor identity trace within same project
Profile fabrication pattern detection
Catches
Professional repeat participants, multi-vendor duplicate entries, identity cycling across deployments
Monitors temporal patterns — completion velocity, inter-keystroke timing, time-of-day clustering, and burst participation patterns — to identify impossible or statistically anomalous response timing that genuine human behaviour cannot produce.
Sub-human completion velocity detection
Burst participation pattern analysis across sessions
Inter-action timing and keystroke rhythm profiling
Catches
Speed-clicking bots, click farm burst sessions, 1,000+ daily attempt patterns, automated completion timing
Executes study-specific business logic — demographic eligibility rules, quota controls, exclusion and inclusion criteria — alongside Friskr's own deduplication signals. Cross-deployment duplicate elimination within the same project is included at the base rate.
Study-specific eligibility and quota enforcement
Cross-deployment deduplication within project scope
Custom exclusion and inclusion criteria execution
Catches
Quota gaming, ineligible respondent fabrication, project-level duplicate re-entry across all vendor sources
Layer 2 · Biometric · Beta
FaceGraph™
Escalation
fires on low Friskore™
Webcam liveness verification that fires only when the composite Friskore™ score falls below the deployment's configured threshold. Provides cryptographic proof of human presence without storing photos or personal data. Privacy-by-design — no biometric data is retained post-verification.
Liveness detection — confirms real human presence vs. photo or screen replay
Uniqueness verification — same person cannot enter multiple times even across devices
Age estimation — demographic cross-validation without ID documents
Deepfake detection — trained on 18M+ real webcam sessions
Catches
Sophisticated human fraud that passes all 8 clusters, identity sharing, account pooling, mask or photo attacks — the last line of defence before a respondent is passed
$0.10 / escalation · from wallet
FaceGraph™ is not included in the $0.90 Per Access or $1.93 Per Pass rates. It is charged from your wallet credits per escalation. Typical escalation rate is 5–10% of passes. At 7% escalation on 1,000 passes, FaceGraph™ adds $7.00 to the study total.
How clusters combine on real fraud scenarios
Each fraud vector triggers multiple clusters simultaneously. The Friskore™ score is the weighted composite of all active signals — a single cluster never decides the outcome alone.
Click farms
BotGraph™ · DeviceGraph™ · TraceGraph™ · ChronoGraph™
+ NetGraph™ support
VPN masking
VirtuGraph™ · NetGraph™ · GeoGraph™
+ TraceGraph™ support
AI bots
BotGraph™ · ChronoGraph™
+ DeviceGraph™ · RuleGuard™ support
Repeat participation
TraceGraph™ · DeviceGraph™ · ChronoGraph™
+ RuleGuard™ support
Botnet geo-spoofing
NetGraph™ · VirtuGraph™ · GeoGraph™
+ BotGraph™ support
Coordinated sabotage
RuleGuard™ · TraceGraph™ · ChronoGraph™
+ NetGraph™ support