Abstract

Data concerning telephone calls is collected by telecom munications operators, typically to support the billing pro cess. However, such records may also be analysed in or der to identify cases of fraudulent usage, which may de pend critically on differentiating normal from abnormal be havioural patterns. Data to support empirical study by the research community within the fraud detection area is hard to acquire, as it contains sensitive information, from both commercial and personal perspectives. Hence, the synthe sis of representative data is a viable way forward, provided reasonable models for telecommunications traffic are avail able. This paper provides such models, based on the be haviour of a sample of 250 subscribers over a 52 week pe riod. This work is partially funded by the EC 6th Frame work IST project Misuse Detection System (IST-026459).