Abstract:

The authors analyzed the intrusion feature selection methods and algorithms based on ReliefF. Combining with the characteristics, in which data points have high similarity in the same class and nonsimilarity in different classes, and optimizing the compute method of intrusion feature weight, we proposed an improved algorithm ReReliefF. It resolved the problems about processing efficiency with the data dimensions in network security. Experiments on the network security test datasets showed the effectiveness of the proposed algorithms, and the improved algorithm has advantages on performance compared to the traditional one.

Key words:  intrusion detection, feature selection, ReliefF algorithm