Computer security problems solving by automatically designed neural network ensembles

Today, computers are becoming more powerful and interconnected that makes their security one of the most important concerns. Conventional security software requires a lot of human effort to identify and work out threats. This human labor intensive process can be more efficient by applying machine learning algorithms. Artificial neural networks are one of the most widely used data mining techniques here. The highly increasing computing power and technology made possible the use of more complex intelligent architectures, taking advantage of more than one intelligent system in a collaborative way. This is an effective combination of intelligent techniques that outperforms or competes to simple standard intelligent techniques. One of the hybridization forms, the ensemble technique, has been applied in many real world problems. In this paper, artificial neural networks based ensembles are used for solving the computer security problems. We apply the self-configuring genetic programming technique to construct symbolic regression formula that shows how to compute an ensemble decision using the component ANN decisions. The algorithm involves different operations and math functions and uses the models providing the diversity among the ensemble members. Namely, we use neural networks, automatically designed with our GP algorithm, as the ensemble members. The algorithm automatically chooses component ANNs which are important for obtaining an efficient solution and doesn ’t use the others. Performance of the approach is demonstrated with test problems and then applied to two real world problems from the field of computer security - intrusion and spam detection. The proposed approach demonstrates results competitive to known techniques. With the approach developed an end user has no necessity to be an expert in the computational intelligence area but can implement the reliable and effective data mining tool.