Parallel Algorithm of Feature Reduction in Intrusion Data

This paper defines the importance of attack features using conditional rough entropy of knowledge and presentes a parallel algorithm of optimal feature selection in intrusion data based on conditional rough entropy. The algorithm divides the decision table of intrusion data into several sub-tables, and then the conditional rough entropy is used for the parallel computing of the sub-tables. Finally, the original decision table reduction is obtained based on the part reduction results from the sub-tables. The proposed algorithm has good performance and is good at dealing with the huge volume of data. The experimental results show that it is effective to reduce the storage requirements of the dataset and the computational cost, and it can increase the detection speed and without sacrificing the detection correctness by using the reduced feature subset.