A new network intrusion detection algorithm: DA‐ROS‐ELM
Yi Yu, Songlin Kang, He Qiu · IEEJ Transactions on Electrical and Electronic Engineering · 2018
In this paper, a novel dual adaptive regularized online sequential extreme learning machine (DA‐ROS‐ELM) is proposed to detect network intrusion. The ridge regression factor based on Tikhonov regularization is introduced to solve the over‐fitting and ill‐posed problems. According to the arrived data in each updating phase and all currently available data, dual adaptive mechanism is designed to respectively select the suitable updating mode of output weightβand regularized parameterC. The performance of our algorithm is assessed by NSL‐KDD dataset, and the results show that the DA‐ROS‐ELM can obtain faster training speed, higher accuracy, lower rate of false positive and false negative, and greater generalization performance than other network intrusion detection algorithms. Besides, the adaptive mechanism makes this algorithm can meet the real‐time requirement of the network intrusion system. © 2018 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.