Detecting mobile telephone misuse

Apparatus for the detection of fraudulent use of a telephone subscriber's instrument in a mobile telephone system, the apparatus including an input preprocessor, a neural network engine coupled to the preprocessor, and an output postprocessor coupled to the neural network engine, wherein the preprocessor is adapted to determine for each subscriber, from that subscriber's telephone call data, a first long term calling profile, a second short term calling profile, and a subscriber profile pattern comprising the difference between the first and second profiles, each said calling profile and subscriber profile pattern comprising a set of values for a respective set of call attributes, wherein the neural network engine comprises a self organising map trained to effect pattern recognition of said subscriber profile patterns and a multilayer perceptron adapted to determine for each recognised pattern a value indicative of the probability of a fraud being associated with that pattern, and wherein said postprocessor is arranged to order said recognised pattern according to said fraud probabilities.

A method for the detection of fraudulent use of a telephone subscriber's instrument in a mobile telephone system, the method including determining a long term calling profile for a said subscriber, determining a short term calling profile for the subscriber, determining the difference between the long term and short term profiles, said difference comprising a subscriber profile pattern, and processing the pattern via a trained neural net arrangement comprising a self organising map adapted to effect pattern recognition of said subscriber profile patterns and a multilayer perceptron adapted to determine said probability value for each recognised pattern whereby to determine from the subscriber profile pattern a probability value for the existence of fraud in that pattern.