Classification model of network intrusion using Weighted Extreme Learning Machine
Worachai Srimuang, Silada Intarasothonchun · 2015
The development of a model classification intrusion detection using Weighted Extreme Learning Machine was examined with KDD'99 data set ad 4 types of main attack : Denial of Service Attack (DoS), User to Root Attack (U2R), Remote to Local Attack (R2L), and Probing Attack, when comparing the effectiveness of working process of the method presented to SVM+GA[6] and ELM, found that weighted technique using RBF Kernel activation function which the value of trade-off constant C was at 25, which was presented the average effectiveness of accuracy to be more effective than other 2 techniques, giving accuracy effectiveness value of DoS = 99.95%, U2R = 99.97%, R2L = 93.64% and Probing = 96.64 %, meanwhile it used less time for working. This could be an interesting technique to be applied to enhance the effectiveness of security of system surveillances in monitoring to be able to remedy the situations on time.