Tag: neural network

This excerpt from my final year project explores different types of telecom fraud, which I'll use as a basis for a C# neural network. Telecom fraud can be categorized into subscription fraud (false identities or payment evasion), call surfing (unauthorized network access through methods like call forwarding or cloning), ghosting (manipulating systems to avoid billing), accounting fraud (internal manipulation of billing systems), and information abuse (misuse of client or system data). I discuss the financial impact of these methods, such as bad debt from subscription fraud and costs incurred from call surfing via PABX manipulation. I also touch upon the vulnerabilities of older analogue mobile phones to cloning and the methods used in ghosting, like tone generating hardware. The project aims to address these fraud types through a neural network approach.

I've shared my Call Detail Record (CDR) Generator, a Microsoft Access 2003 application I developed for my final year project. This tool creates thousands of simulated phone calls, mimicking various customer profiles (e.g., high/low usage, fraudulent, national rate). It offers extensive customization options for call cost, duration, and user behavior patterns (e.g. business vs. home user calling time). The generated call data is designed for training neural networks. Feel free to download and experiment – it's easy to use.