Combining real-time BSS data with CDP data and 3-rd party data to prevent Roaming Fraud, International Revenue Share Fraud (IRSF). Pipeline In -> Analytics + Search -> Push out to Kafka (future) -> Service action + Push out to Data Lake
- Remarks
- For simplicity currently there is no CDC-In from MySQL
- CDC-In has been done for Mongo via Kai for User Info
- CDC-Out has not been implemented due to time constraints
- Cell Towers
- Data generated from the towers - stored in an S3 bucket as
- DDL
- SIM ID
- Location
- Usage Type (Call, Data, Text)
- Data Transfer Amount
- Call Duration
- Timestamp
- DDL
- Data generated from the towers - stored in an S3 bucket as
- Helios
- Pipeline from Cell Tower S3
.csvfiles - CDC-In (Mongo Kai) for User Table
- Copy to new Table for new - Full text search v2 (JLucene + JSON full text index)
- Fraud Detection Algo
- Notebook for Rapid Testing
- Deployed using a Procedure
- Cases
- Impossible Travel - Same SIM ID within a short amount of time registered to Towers that are too far apart
- New record in Fraud detected Table
- Pipeline from Cell Tower S3
- UI
- Dashboard showing the detected Frauds on a Map
- Full text search v2 (JLucene + JSON full text index) Query
- Not covered yet
- Schema Pipeline Inference
- DB Branching
- Ingest from Iceberg
- Auto scaling / fast scaling
- CDC-Out
Click the Demo for better quality
Demo Workspace
- City Base Data
- SIM IDs Base Data
- Data Generation Script to create .csv Files
- Data Generation Script to create fake Users and write to Mongo
- Clone this repository
- Build Docker Container
docker buildx build --no-cache -f ./Dockerfile -t singlestore-demo/telco-fraud .
- Create a
.envfile fromenv.sample - Run the Container
docker run -p 3000:3000 --env-file .env -d --name telco-fraud singlestore-demo/telco-fraud - Open your browser and point it to localhost:3000