Graph Gurus Episode 34: Graph Databases are Changing the Fraud Detection and Prevention Landscape for Financial Institutions
0 likes133 views
Episode 34 of Graph Gurus discusses how graph databases are transforming fraud detection in financial institutions, highlighting the significant financial losses and inefficiencies in traditional fraud detection methods. The session emphasizes the advantages of using TigerGraph's technology to reduce false positives, automate processes, and enable real-time fraud detection. Key use cases are presented, showcasing the capabilities of graph databases to uncover complex relationships among fraudulent activities.
Graph Gurus Episode 34: Graph Databases are Changing the Fraud Detection and Prevention Landscape for Financial Institutions
- 1. Graph Gurus Episode 34 Graph Databases are Changing the Fraud Detection and Prevention Landscape for Financial Institutions
- 2. © 2020 TigerGraph. All Rights Reserved Today's Host 2 David Ronald Director of Product Marketing ● Prior work in artificial intelligence, natural linguistic programming and telecommunications technology ● 20+ years in tech industry ● BSc in Applied Physics from Strathclyde University, MSc in Optoelectronic & Laser Devices from St Andrews
- 3. © 2020 TigerGraph. All Rights Reserved Today's Presenter 3 Abhishek Mehta Director of Sales Engineer ● McKinsey, Bloomberg, Cisco & Dabizmo (NLP Startup) Founder ● 15+ years designing and implementing complex analytics solutions for Fortune 100 companies ● Patents in NLP spanning Conceptunary Ontology Design, Language Pattern Recognition, and Conversion
- 4. © 2020 TigerGraph. All Rights Reserved Some Housekeeping Items ● Although your phone is muted we do want to answer your questions - submit your questions at any time using the Q&A tab in the menu ● The webinar is being recorded and will uploaded to our website shortly (https://www.tigergraph.com/webinars/) and the URL will be emailed you ● If you have issues with Zoom please contact the panelists via chat 4
- 5. © 2020 TigerGraph. All Rights Reserved Key Stats on Credit Card Fraud (2018) 5 ● Losses of around $25 billion from payment fraud worldwide ● Credit card losses in the U.S. alone are ⅓ of the total ● 33 million adults, around 15% of US cardholders had blocked sales - blocked sales amounts to $118 billion, while the cost of real card fraud only amounts to $8-9 billion ● Only 1 in 5 blocked transactions are actually fraudulent - 40% of denied users are attempting to pay over $250 for a sale
- 6. © 2020 TigerGraph. All Rights Reserved Framework: Total Cost of Fraud 6
- 7. © 2020 TigerGraph. All Rights Reserved Think Beyond Fraud Rate 7
- 8. © 2020 TigerGraph. All Rights Reserved Insurance Fraud Telecom Fraud Financial Fraud Health/Parma Others Worker Compensation Device Theft Credit Card Fraud Claims/Billing Licensing Fraud Car, Home Fraud Spam Call Detection Counterparty Risks Prescription Rideshare Fraud Fraud Detection with Graph is Everywhere AML 8
- 9. © 2020 TigerGraph. All Rights Reserved Relational Database Key-Value Database Graph Database Customer XXXXXX Product XXXXXXXX Supplier XXXXXXXX Location XXXXXXXX Order XXXXXXXXX Product Customer Supplier Location KEY VALUE XXXXX Order Customer Prod uct Supplier • Rigid schema • High performance for transactions • Poor performance for deep analytics • Highly fluid schema/no schema • High performance for simple transactions • Poor performance deep analytics • Flexible schema • High performance for complex transactions • High performance for deep analytics Location 2 Product Payment PURCHASED RESIDES SHIPSTO PURCHASED SHIPS FROM A C C EPTED MAKES The Evolution of Databases XXXXX XXXXX XXXXX Location 1 N O TIFIES 9
- 10. © 2020 TigerGraph. All Rights Reserved Why TigerGraph For Fraud - Reason 1 10 Total Cost of Fraud = Fraud losses + Tools & head count + CLV impact TigerGraph reduces false positives TigerGraph finds fraudulent patterns buried deep in data TigerGraph reduces manual fraud processing overhead
- 11. © 2020 TigerGraph. All Rights Reserved Core Technology Groups Machine Learning, Data Science Groups, Enterprise Architecture, Infrastructure 3 Fraud Modelling Group Statisticians, Domain Experts, Play/Rule books, Experts in Credit, Fraud Behaviour 2 Operational Groups Fraud Analyst, Compliance , Regulations, Forecasting, Reporting, Visualizatoins 1 Why TigerGraph For Anti-Fraud Strategy - Reason 2 Detect Prevent Respond 11
- 12. © 2020 TigerGraph. All Rights Reserved Core Technology Groups Feature extraction; reputation list 3 Fraud Modelling Group Graph AI/ML based feature extraction; various statistical functions 2 Operational Groups Rule-based fraud detection; visualization 1 Demonstration: AML Detect Prevent Respond 12
- 13. © 2020 TigerGraph. All Rights Reserved The Art of the Possible Use Case : Expand the List of Probable Fraudsters Process from Last 20 Years Two, Three hop queries : Process Step 1: Select all known Fraudsters export them into Excel Sheet/SQL table. Step 2: Select PII attributes (Name, locations, address, email…). Step 3: Search/Select by fraudster attributes 1-by-1 or as a collection to get the possible overlapping accounts. Step 4: Manually, review the newly found accounts to flag as fraudsters. The Future is Now Five, Six, 10+ hop queries Step 1: For fraudster traverse PII Edges. Hop to the other accounts that link to the PII edges. Step 2: Execute “similarity detection” on PII nodes; over 0.9 threshold is fraudsters over 0.5 is candidate for review. ● Seconds vs T+Days ● Real-time execution as soon as a new fraudster is loaded. Don’t wait for critical mass or weekly effort. ● Automation: Tribal knowledge of the fraud-detection ● Reduce false positives with high thresholds 13
- 14. © 2020 TigerGraph. All Rights Reserved Expensive - High TCO Two, Three hop queries : Process Step 1: Select all known Fraudulent Transactions Step 2: Select all touchpoints of these transactions (devices, ip addrs, Name, locations, address, email…). Step 3: Inner Join to get one hop device/payment inst. Details. Continue as many hops you need to do. Step 4: Stitch the results together. The Future is Now Five, Six, 10+ hop queries Step 1: Shortest Path Algorithm/All Paths/Weighted ● Seconds vs T+Days ● Real-time execution as soon as a new fraudster is loaded. Don’t wait for critical mass or weekly effort. ● Automation: Tribal knowledge of the fraud-detection ● Reduce false positives with high thresholds The Art of the Possible Use Case : All Possible Links to Fraudulent Transactions 14
- 15. © 2020 TigerGraph. All Rights Reserved Rare Implementation Step 1: Use auto/feature extractors based on various attributes. Step 2: Clustering Step 3: Order The clusters based on number of frauds in each cluster. Step 4: Manually, review or some some automation. The Future is Now Five, Six, 10+ hop queries Step 1: Community detection based on edges/attributes Step 2: PageRank the nodes/credit card Step 3: Pick up communities with highest/average page ranks The Art of the Possible Use Case : Unsupervised Learning - Fraud Ring Detection Credit Cards 15
- 16. © 2020 TigerGraph. All Rights Reserved Core Technology Groups Feature extraction; reputation list 3 Fraud Modelling Group Graph AI/ML based feature extraction; various statistical functions 2 Operational Groups Rule-based fraud detection; visualization 1 Demonstration: AML Detect Prevent Respond 16
- 17. © 2020 TigerGraph. All Rights Reserved Lessons from Top Implementations 17 MPP ● GSQL - Intuitive, Turing complete graph query language for developing complex analytics in days ● User extensible graph algorithms library TCO ● Less Hardware ● High QPS Enterprise Grade Security ● Encryption Support ● Control access to sensitive data based on user role, dept or organization with MultiGraph Real-time Performance Sub-second response for queries touching tens of millions of entities/relationships Scalability for Massive Datasets 100 billion+ entities, 1 trillion+ relationships Deep Link Multi-Hop Analytics Queries traverse 10+ hops deep into the graph performing complex calculations
- 18. Q&A Please submit your questions via the Q&A tab in Zoom
- 19. © 2020 TigerGraph. All Rights Reserved More Questions? Join our Developer Forum https://community.tigergraph.com Join our Developer Chat https://discord.gg/F2c9b9v Sign up for our Developer Office Hours (Thursday at 11 AM PDT) https://info.tigergraph.com/officehours 19
- 20. © 2020 TigerGraph. All Rights Reserved Additional Resources Start Free at TigerGraph Cloud https://www.tigergraph.com/cloud/ Test Drive Online Demo https://www.tigergraph.com/demo Download the Developer Edition https://www.tigergraph.com/download/ Guru Scripts https://github.com/tigergraph/ecosys/tree/master/guru_scripts 20
- 21. © 2020 TigerGraph. All Rights Reserved Upcoming Events Graph Gurus 35: How the New Functionality in 3.0 will Help You Connect the Dots Even Better Wednesday, May 20, at 11am PDT https://info.tigergraph.com/graph-gurus-35 Graph Gurus 35: How the New Functionality in 3.0 will Help You Connect the Dots Even Better 21 Graph Gurus Comes To You Workshop with Galvanize Thursday, May 14, 4:30-6:30 pm PDT https://info.tigergraph.com/graph-gurus-comes-to-you -galvanize
- 22. Thank You