Abstract: A method of and system for detecting the possible fraudulent use of a telecommunications network involves applying rule-based criteria to generate a plurality of fraud alarms, each corresponding to an individual rule. Each alarm is associated with a particular customer, and for each individual customer a note is made of the total alarms generated by that customer and the grouping of individual alarm types generated. The customer's call is then determined to be fraudulent or otherwise based upon prior experience of past customers who have generated that particular profile of alarm grouping and total number of alarms. The system automatically outputs a list of potentially fraudulent customers, the accounts of which may either be further investigated or may automatically be inhibited.