Fraud Detection Demo

To execute the fraud detection demo locally using netcat as the data source, follow these steps:

1. Start a netcat listener on port 9999:

   nc -lk 9999

2. Run the main method of the class com.ververica.field.dynamicrules.Main in your IDE or via CLI.
3. Send rule updates or control commands to the netcat listener in the required format.

The React application is configured to orchestrate both the Java backend and the React development server. First, install the required dependencies, then start the application.

npm install
npm start

Navigate to the demo-backend directory and use Docker Compose to start the Kafka services. The log level is set to CRITICAL to minimize noise.

cd demo-backend
docker-compose --log-level CRITICAL up

Use Maven to clean, install dependencies, and run the Spring Boot application.

mvn clean install spring-boot:run

The dashboard visualizes the flow from Transactions $\rightarrow$ Rules $\rightarrow$ Alerts.

• Rules: Fetched from /api/rules. Each rule is associated with a specific transaction.
• Alerts: Received via WebSocket. An alert is linked to a rule via the ruleId property.
• Visual Logic:
  • Lines connect Transactions to Rules.
  • Lines connect Rules to Alerts.
  • If a rule is currently triggering an alert, the connection line color changes to red (#dc3545); otherwise, it remains orange (#ff7f50).

The project uses a docker-compose.yaml file to orchestrate a fraud detection data pipeline consisting of Zookeeper, Kafka, Schema Registry, and a Schema Registry UI.

To start the entire stack, run:

docker-compose up

Service Overview

• Zookeeper (zoo1): Manages Kafka cluster state. Accessible on port 2181.
• Kafka (kafka1): The message broker. It is configured with two listeners:
  • LISTENER_DOCKER_INTERNAL: Used for communication within the Docker network (port 19092).
  • LISTENER_DOCKER_EXTERNAL: Used for external access from the host machine (port 9092).
• Schema Registry (schema-registry): Manages Avro schemas. Accessible on port 18081.
• Schema Registry UI (schema-registry-ui): A web interface to view schemas. Accessible on port 8001 (mapped from container port 8000).

The backend service is a Spring Boot application. You can start the application by executing the main method in the com.ververica.demo.backend.Main class. This initializes the Spring context, enables auto-configuration, and performs a component scan across the com.ververica package to wire up the necessary beans for the fraud detection demo.

package com.ververica.demo.backend;

import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.EnableAutoConfiguration;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.context.annotation.ComponentScan;

@SpringBootApplication
@EnableAutoConfiguration
@ComponentScan("com.ververica")
public class Main {
  public static void main(String[] args) {
    SpringApplication.run(Main.class, args);
  }
}

The UI includes a slider to control the speed of the transaction generator. Changing the slider value updates a local storage key generatorSpeed and triggers a fetch request to the backend to adjust the simulation speed.

Note that the value sent to the API is transformed by getFakeValue(), which scales the slider input (0-30) into a higher magnitude value used by the generator.

You can send special control commands to the system using JSON format to manage the state of the fraud detection engine. The ruleState must be set to "CONTROL".

Available controlType values:

• DELETE_RULES_ALL: Deletes all rules.
• EXPORT_RULES_CURRENT: Exports the current rules.
• CLEAR_STATE_ALL: Clears all internal state.

Examples:

{"ruleState": "CONTROL", "controlType":"DELETE_RULES_ALL"}
{"ruleState": "CONTROL", "controlType":"EXPORT_RULES_CURRENT"}
{"ruleState": "CONTROL", "controlType":"CLEAR_STATE_ALL"}