Skip to content

Latest commit

 

History

History
211 lines (163 loc) · 7.6 KB

File metadata and controls

211 lines (163 loc) · 7.6 KB

🔍 Fraud Investigation Notes

Aleksandr Ovchinnikov · Fraud Analyst · OSINT Specialist
📧 alexrmovch@gmail.com · 🔗 LinkedIn


🎯 Portfolio Overview

Professional fraud investigation portfolio demonstrating end-to-end investigation methodology, SQL pattern detection, and OSINT intelligence gathering through detailed case studies.

Total Fraud Prevented: $25,000 (synthetic data)
Cases Completed: 6 investigations
Average Detection Time: 1.6 hours
SQL Query Precision: 98.7% (1.8% false positive rate)

Note

Disclaimer: All data in this repository is synthetic and created solely for demonstration purposes to showcase investigative methodology and technical skills.


🔥 Featured Investigation

Case #1: Mass Registration Fraud Ring

Alert: 47 new accounts from single IP in 2 hours
Investigation Time: 47 minutes
Outcome: Discovered coordinated fraud ring with 1,247 accounts
Fraud Prevented: $5,000 (synthetic)

Key Findings:

  • ❌ Same device fingerprint across all registrations (bot automation)
  • ❌ Email domain registered 3 days before attack (disposable)
  • ❌ Form submission time: 1.8 seconds average (bot speed vs human 15-45s)
  • ❌ All transactions attempted to same merchant with stolen cards

Detection Method: Used SQL velocity clustering combined with device fingerprint correlation and OSINT domain analysis to identify the fraud ring within 47 minutes.

📄 Read Full Investigation →

Fraud Ring Network
Network visualization of 47-account fraud ring linked by shared device fingerprints


📊 All Case Studies

# Investigation Type Fraud Prevented Detection Time Status
#1 Registration Fraud Ring $5,000 47 min ✅ Complete
#2 Credential Stuffing → ATO $2,000 1.2 hrs ✅ Complete
#3 Impossible Travel Detection $1,500 23 min ✅ Complete
#4 Bot Network Identification $15,000 3.5 hrs ✅ Complete
#5 Session Hijacking $500 15 min ✅ Complete
#6 Email Domain Fraud Ring $1,000 2.1 hrs ✅ Complete

🎯 Detection Performance

SQL-based fraud detection rules performance on synthetic datasets:

Detection Method Precision Recall False Positive Rate Use Case
Bulk Registration Detection 98.7% 91.2% 1.8% Bot network identification
Device Fingerprint Clustering 96.8% 88.4% 3.2% Account linkage analysis
Impossible Travel Detection 99.5% 76.5% 0.5% Geographic anomaly detection
Credential Stuffing Indicators 94.2% 83.1% 2.1% ATO prevention

Industry Benchmark: 5-10% false positive rate
Portfolio Achievement: <3.2% across all detection methods


💡 SQL Expertise Example

Business Challenge: Detect bot networks creating multiple accounts from same infrastructure

My Solution: SQL query clustering suspicious registrations by IP velocity and device sharing:

-- Production query from Case Study #1: Bulk Registration Detection
-- Detects: 5+ accounts from same IP with 3 or fewer unique devices
-- Execution time: ~2.3 seconds on 10K user dataset

WITH registration_velocity AS (
    SELECT 
        registration_ip,
        COUNT(DISTINCT user_id) AS accounts_created,
        COUNT(DISTINCT device_fingerprint) AS unique_devices,
        AVG(EXTRACT(EPOCH FROM form_submission_time)) AS avg_form_seconds
    FROM users
    WHERE registration_timestamp > NOW() - INTERVAL '24 hours'
    GROUP BY registration_ip
)
SELECT 
    registration_ip,
    accounts_created,
    unique_devices,
    ROUND(avg_form_seconds, 2) AS avg_form_time,
    CASE
        WHEN accounts_created >= 50 THEN 'CRITICAL'
        WHEN accounts_created >= 20 THEN 'HIGH'
        WHEN accounts_created >= 10 THEN 'MEDIUM'
        ELSE 'LOW'
    END AS risk_level
FROM registration_velocity
WHERE accounts_created >= 10 
  AND unique_devices <= 3
  AND avg_form_seconds < 5  -- Bot speed threshold
ORDER BY accounts_created DESC;

Business Impact:

  • Detected 1,247-account fraud ring in 47 minutes
  • 98.7% precision with only 1.8% false positives
  • Prevented $5,000 in fraudulent transactions (synthetic)
  • Created automated detection rule now running every 15 minutes

📁 View All 10 SQL Queries →


🔍 OSINT Investigation Capabilities

Demonstrated OSINT methodology across all investigations:

Email Intelligence:

  • Domain age validation (WHOIS)
  • MX record verification (DNS)
  • Disposable email detection
  • Breach correlation (Have I Been Pwned)
  • 87% time reduction through automation (15min → 2min)

IP Geolocation Analysis:

  • VPN/datacenter detection
  • Geographic clustering
  • Abuse reputation scoring
  • Impossible travel calculation
  • 90% automation rate

Device Fingerprint Correlation:

  • Cross-account device linkage
  • Bot signature detection
  • Hardware consistency analysis
  • Network graph construction

📖 View OSINT Playbooks →


📸 Investigation Visualizations

Registration Velocity Timeline
Timeline
Spike in bot registrations detected within 2-hour window on Jan 15, 2026

Fraud Ring Network Graph
Network Graph
Visualization of account linkages through shared device fingerprints and IP addresses

🖼️ View All Visualizations →


🛠️ Technical Skills Demonstrated

Data Analysis:

  • Advanced SQL (PostgreSQL): CTEs, window functions, clustering
  • Python: pandas, numpy, scikit-learn, networkx
  • Data visualization: matplotlib, seaborn, plotly
  • Statistical analysis: anomaly detection, pattern recognition

Fraud Domain Expertise:

  • Registration fraud patterns (bot networks, mass creation)
  • Account takeover detection (credential stuffing, session hijacking)
  • Behavioral anomaly identification (impossible travel, velocity abuse)
  • OSINT investigation methodology

Business Analysis:

  • Risk scoring frameworks (0-100 scale)
  • False positive optimization (<3% target)
  • Business impact quantification ($25K prevented)
  • Operational efficiency metrics (1.6 hour avg detection)

📂 Repository Structure

frinv-notes/
├── Portfolio_Playbook.md  # Master structural guide
├── case_studies/          # 6 detailed investigation reports
├── sql_queries/           # 10 production-ready SQL detection queries
├── osint_playbooks/       # Step-by-step OSINT investigation methodologies
├── analysis_notebooks/    # Jupyter notebooks with data analysis & ML
├── visualizations/        # Charts, graphs, network diagrams
├── methodology/           # Investigation frameworks & best practices
└── metrics/               # Detection performance & business impact tracking