Hugging Face's logo

Datasets:
AskFirstAI
/
global-web-antifraud-behavioral-telemetry

Tasks:
Text Classification
Tabular Classification
Modalities:
Tabular
Text
Formats:
csv
Languages:
English
Russian
Spanish
Size:
< 1K
Tags:
cyber-security
ad-fraud
behavioral-biometrics
clickstream-analytics
bot-detection
human-intent
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
global-web-antifraud-behavioral-telemetry / README.md
AskFirstAI's picture
AskFirstAI
Update README.md
4c99fc1 verified
preview code
|
Raw History Blame Contribute Delete
7.27 kB
metadata 
license: cc-by-nc-4.0
task_categories:
  - text-classification
  - tabular-classification
language:
  - en
  - ru
  - es
  - zh
  - ar
tags:
  - cyber-security
  - ad-fraud
  - behavioral-biometrics
  - clickstream-analytics
  - bot-detection
  - human-intent
  - telemetry
  - anti-fraud
size_categories:
  - 500k<n<1M
pretty_name: AskFirst Global Web Anti-Fraud & Biometrics

AskFirst Global Web Anti-Fraud, Bot Detection & Behavioral Biometrics Dataset (Premium Sample)

1. DATASET OVERVIEW

This repository contains a premium production sample of the Global Web Anti-Fraud & Behavioral Biometrics Telemetry dataset managed by Ask First AI, Inc. This database contains 0% synthetic or AI-generated rows. Every single record represents raw, ground-truth human traffic or malicious automated bot interactions aggregated from live production environments via native Next.js Cloud SDK edge nodes.

The dataset applies a strict Zero-NULL policy, ensuring that all telemetry, technographic, and biometric fields are 100% populated across all rows. To maintain absolute compliance with global privacy standards, all client device signatures, canvas fingerprints, and network subnets are systematically hashed using cryptographic SHA-256 with a secure salt.

  • Total Baseline Volume: 605,915+ fully labeled, verified sessions (dynamically ingested via real-time SDK cloud pipelines).
  • Delivery Engine: Secured via Google BigQuery Analytics Hub Data Clean Rooms (Authorized Views).
  • Target Table Address: gen-lang-client-0835445607.firestore_export.antifraud_behavioral_sessions
  • Licensing Model: Commercial Enterprise Licenses are handled directly via corporate Proforma Invoices or Private Cloud Offers.
  • Geographical & Cross-Lingual Scope: 120+ countries, 5 core human languages detected (English, Russian, Spanish, Chinese, Arabic).

2. DATA SCHEMA SPECIFICATION

Every row inside this database maps an entire synchronized user-to-network tracking lifecycle containing 22 precision fields optimized for standard Google BigQuery SQL environments:

  1. sync_timestamp (TIMESTAMP): Exact ISO-8601 UTC time of session validation and log insertion window.
  2. session_id (STRING): Anonymized unique session token hashed via SHA-256 (sess_<20-char-hex>).
  3. user_question (STRING): Raw, organic human search query or intent string. 100% scrubbed of any PII at edge nodes.
  4. topic (STRING): Categorical structural domain label of the intent query (finance, technology, health, lifestyle, science, business, general).
  5. language (STRING): ISO 639-1 language code detected from client browser negotiation headers (en, ru, es, zh, ar).
  6. country_code (STRING): Country/region of the session origin resolved via secure edge cloud lookup (e.g., United States, European Union, GB, IN, NG).
  7. visit_status (STRING): Definitive Ground-Truth anti-fraud label: approved (verified real human) or fraud (automated bot / click-farm / script execution).
  8. dwell_time_ms (INT64): Confirmed active user dwell time recorded on the page interface, measured in milliseconds.
  9. behavioral_entropy (FLOAT64): Calculated Shannon behavioral entropy mapping interactivity randomness (ranging from 0.0000 for programmatic scripts to 1.0000 for chaotic human movements).
  10. behavioral_data (JSON/STRING): Comprehensive kinematic session profile map capturing interaction counts: {"mouseMovements", "clicks", "scrollEvents", "keyPresses", "timeOnPage", "entropy"}.
  11. user_agent (STRING): Full raw HTTP User-Agent string outlining client browser, engine version, and operating system architecture.
  12. canvas_fingerprint_hash (STRING): 64-character SHA-256 cryptographic hash of the hardware-rendered HTML5 Canvas device fingerprint.
  13. ip_subnet_hash (STRING): 64-character SHA-256 cryptographic hash of the masked network client subnet.
  14. is_proxy_or_tor (BOOLEAN): Security flag indicating whether the session bypassed via a known Tor node, anonymous proxy, or data-center VPN pool (true / false).
  15. turnstile_verified (BOOLEAN): Verification status confirming successful passing of the invisible cryptographic Cloudflare Turnstile challenge (true / false).
  16. honeypot_triggered (BOOLEAN): Intrusion flag signaling if the visitor interacted with invisible, hidden DOM honeypot trap elements (true / false).
  17. trust_score (FLOAT64): Aggregated analytical confidence rating evaluating session legitimacy, scaling from 0.0 (Absolute Malicious Fraud) to 100.0 (Fully Verified Human).
  18. trust_level (STRING): Categorized operational trust tier derived from the scoring layer (low, medium, high, verified).
  19. referral_fraud_rate (FLOAT64): Historical running fraud ratio associated with the specific traffic source referrer (scaled 0.0 to 1.0).
  20. visit_fraud_rate (FLOAT64): Historical ratio of confirmed fraudulent visits linked to the ingestion endpoint (scaled 0.0 to 1.0).
  21. verified_clicks (INT64): Counter tracking confirmed, biologically validated human navigation clicks performed within the session lifecycle.
  22. is_auto_verified (BOOLEAN): Real-time flag indicating if the session was immediately validated on-the-fly by biometric edge heuristics (true / false).

3. PRIMARY TARGET USE CASES

  • Anti-Fraud & WAF ML Training: Ideal for building binary classification algorithms (visit_status) and real-time risk scoring matrices (trust_score) to block advanced scrapers and credential-stuffing botnets.
  • Behavioral Biometrics Research: Provides deep kinematics data (mouse arcs, keystroke flight times, and Shannon entropy coefficients) across multiple desktop and mobile operating systems.
  • Botnet & Click-Farm Clustering: Allows cybersecurity analysts to map malicious nodes and device-spoofing networks by cross-referencing canvas_fingerprint_hash + ip_subnet_hash + user_agent patterns.
  • Organic Search Intent Profiling: Offers clean data on what real, verified humans are searching for online, completely filtered from click-inflation and bot noise.

4. ENTERPRISE COMMERCIAL LICENSING & SUBSCRIPTIONS

Delivery is executed instantly via secure Cloud Clean Rooms inside the Google BigQuery Analytics Hub. Corporate merchant routing is cleared via PayPal Business US through formal Proforma Invoices:

  1. Historical Production Archive (One-time Bulk Purchase): €5,500
    Immediate uncapped read access to the complete baseline database of 605k+ records.
  2. Small Batch Continuous Ingestion Subscription: €800 / month
    Automated streaming delivery up to 10,000 fresh verified rows per month.
  3. Enterprise Ingestion Subscription: €2,500 / month
    Automated streaming delivery up to 50,000 fresh verified rows per month.
  4. Unlimited Real-Time Ingestion Stream: €4,500 / month
    Uncapped, live streaming directly from our edge SDK pipeline into your corporate SIEM, fraud-prevention, or model training architecture.

5. CONTACT & DATA ROOM ENQUIRIES

  • Corporate Infrastructure Hub: Ask First AI, Inc. (United States)
  • Primary Communications Desk: boss@askfirst.ai