Real Time Fraud Detection with Stateful Functions
summary
Real‑time fraud and bot detection for webshops implemented with Flink Stateful Functions, Kafka/MSK, Redis and Postgres, running on EKS; HTTP logs plus optional tracking feed into a session aggregator that calls Python session evaluators over HTTP using heuristics and ML to score every session and map suspicious events to unique transaction IDs. Architecture supports zero downtime deployment of new rules/models, exposes JSON outputs and a back‑office UI, and proves viable in production with tuning (use SSDs for hot state).
AI-written summary · about
chapters (13)
- 0:00Introduction
- 6:41Why Flink
- 9:03Architecture
- 11:48Components
- 15:38Fraud Back Office
- 17:07Fraud Attackers
- 19:18Anomalies
- 20:22Attack
- 21:08Challenge
- 22:02Aggregate Session
- 23:15Classification
- 25:58Mitigation
- 27:06Lessons
related talks
25mReal-Time Fraud Detection at Scale
13mBuilding Financial Fortresses: Real-Time Fraud Detection
15mRealtime bot detection with Flink
51mReal-time Fraud Detection: Challenges and Solutions
10mReal-Time Fraud Prevention at Scale
29mBuilding Resilient AI Platforms: Financial Fraud Detection
16mAnalysing Real time Activity Streams for Security
13mBehavioral Analytics for Real-Time Fraud Prevention