Machine Learning Module for Fraud Detection

This technical report outlines the development of a machine learning-based module designed to detect fraudulent activities within a web portal used by a major telecom company in Russia. The objective was to create a protection system capable of distinguishing between normal user actions and various types of fraudulent behavior, which can include access from fake IP addresses and credit card theft attempts. The report details the project stages, including the analysis of potential threats, implementation of machine learning algorithms, and the development of an administration module for user interface management. The system employs fuzzy logic to enhance fraud detection capabilities and is supported by a RESTful API for integration. Extensive testing procedures were implemented to ensure the module's reliability and performance, including load testing and functional testing using automated tools. The report concludes with a description of the project's outcomes, including the successful deployment of the Anomaly Monitor subsystem, which is currently in trial mode with the client.