A Victim-Based Framework for Telecom Fraud Analysis: A Bayesian Network Model

Abstract

The increasingly rampant telecom network fraud crime will cause serious harm to people's property safety. The way to reduce telecom fraud has shifted from passive combat to active prevention. This paper proposes a victim analysis and prediction method based on Bayesian network (BN), which models victims from age, gender, occupation, marriage, knowledge level, etc. We describe the fraud process in terms of whether to report to the police, property loss, and realizing the reasoning of the whole process of telecom fraud. This paper uses expert experience to obtain a Bayesian network structure. 533 real telecom fraud cases are used to learn Bayesian network parameters. The model is capable of quantifying uncertainty and dealing with nonlinear complex relationships among multiple factors, analyzing the factors most sensitive to property damage. According to the characteristics of victims, we conduct situational reasoning in the Bayesian network to evaluate property damage and alarm situations in different scenarios and provide decision support for police and community prevention and control. The experimental results show that male staff in government agencies are the most vulnerable to shopping fraud and women in schools are the most vulnerable to phishing and virus fraud and have the greatest property loss after being deceived; victim characteristics have very limited influence on whether to report to the police.