Data-Driven Artificial intelligence (AI)-based System for Identifying Fraudulent Telephony Activities

Abstract

The telecommunications industry is an essential part of the modern world of communication tools as it links billions of users across the globe. Nevertheless, telephony fraud has become a widespread issue, and it includes an assortment of fraud types, including subscription fraud, SIM cloning, toll fraud, and voice phishing (vishing), which cost people significant amounts of money and compromised the credibility of the service. The traditional number-based fraud detection techniques feel restricted by the dynamic strategies used by the fraudsters such as spoofing of caller IDs and spoofing of numbers, thereby causing high rate of false positive and late response. This paper suggests a machine-driven Artificial Intelligence (AI) model to detect suspicious telephony transactions with a big amount of textual information. The Baidu search engine was used to gather textual information about actual telecom fraud cases, which was preprocessed by removing noise, segmenting, and tagging part-of-speech and converted into TF-IDF feature vectors. This data was trained in an Artificial Neural Network (ANN) to learn complicated patterns of fraud. Measures of evaluation such accuracy, precision, recall, and F1-score indicate a strong performance of the system with an accuracy of 98.53. The findings indicate that AI-based models have the potential to identify the constantly changing telephony fraud, which is scalable and can deliver timely information to defend users and service providers. It is a new and innovative adaptive scheme of telecom fraud detection, as opposed to the conventional schemes.