Behavior-Based Intrusion Detection in Mobile Phone Systems

The field of mobile and wireless networking is reemerging amid unprecedented growth in the scale and diversity of computer networking. However, further increases in network security are necessary before the promise of mobile communication can be fulfilled. In this paper, we describe how neural networks and tools can be applied against frauds in analog mobile telecommunication networks. To the best of our knowledge there has been a lot of work done to secure the usage of mobile phones at the hardware level, but very little at the software level. In this paper, we present an on-line security system for fraud detection of impostors and improper use of mobile phone operations based on a neural network (NN) classifier. It acts solely on the recent information and past history of the mobile phone owner activities, and classifies the telephone users into classes according to their usage logs. Such logs contain the relevant characteristics for every call made by the user. As soon as the system identifies a fraud, it notifies both the carrier telecom and the victim about it immediately and not at the end of the monthly bill cycle. In our implementation, we make use of the radial basis function (RBF) model because of its simplicity and its flexibility to adapt to pattern changes, i.e., it encompasses the important characteristic of learning. By learning, a RBF NN can discover some regular patterns and the relation across them, and organize itself for making these associations. As a consequence it is widely used for solving classification and pattern recognition problems. Our results indicate that our system reduces significantly the telecom carriers's profit losses as well as the damage that might be passed to the clients. This might help the carriers to reduce the cost of phone calls and will, in turn, benefit the users.