Fortifying Digital Trust: Next-Gen Strategies in Fraud Prevention
summary
Describes a production system that combines Neo4j graph database with LLMs using RAG + vector search to turn implicit signals into explicit graph patterns, running real‑time checks at click‑to‑pay. Uses 1.6 million user nodes and ~400,000 purchase events to detect collusive merchant rebate fraud, reduces false positives and increases detection throughput (claims a 300% improvement for a banking infra customer), and delivers knowledge‑graph ingestion pipelines that go from data to live detection in weeks.
AI-written summary · about
chapters (14)
- 0:00Introduction to fraud prevention
- 1:43Impact and scope of fraud
- 4:11Limitations of traditional fraud systems
- 8:45Graph technology introduction
- 13:23Graph machine learning for fraud
- 15:03Real-time fraud detection example
- 17:52Graph database integration
- 19:17Generative AI and fraud evolution
- 22:28LLMs for knowledge graphs
- 24:58Scaling fraud prevention systems
- 25:19Case study speed improvement
- 26:22Case study accuracy enhancement
- 27:05Case study coverage expansion
- 27:49Conclusion and community resources
related talks
16mLeveraging ML to Create a Reliable Anti-Fraud System
35mGraphRag in Action: Revolutionizing Fraud Detection Techniques
25mReal-Time Fraud Detection at Scale
12mHarnessing AI & LLMs for Next-Gen Digital Payment Security
13mHarnessing Machine Learning to Combat E-commerce Fraud
26mAI-Powered Fraud Detection in Cloud-Native Digital Payments
13mBehavioral Analytics for Real-Time Fraud Prevention
45mAnti-Fraud in Reverse and the Use of ML Methods in It