An Effective Deep Learning Algorithm for Intrusion Detection

EasyChair Preprint 3452

Abstract

Although many mature technologies can be used to prevent cyberattacks, we still have to redesign or adjust these detection or defense systems for unknown cyberattacks today.
In fact, the speed of update for these systems may be slower than the production of cyberattacks.
To solve this problem, the technology of intrusion detection and prevention system (IDPS) is indispensably required for network because it is capable of not only detecting the unknown attacks but also preventing attacks.
More recently, using deep learning technology for the IDPS to precisely detect the attacks in a network is a promising research topic.
However, because the parameter setting in deep learning still relies on manual operation by human, setting parameters, such as the number of layers for a neural network and the number of neurons in each layer, to reach a higher accuracy still is a critical issue in this research domain.
In this paper, we will present an effective algorithm, which combines a metaheuristic algorithm with deep learning to dynamically adjust the parameters of deep learning (i.e., the number of neurons in each hidden layer) to enhance the performance of deep learning for IDPS.
The experimental results show that the accuracy of our proposed method is outperform than current deep learning methods compared in this research for IDPS.

Keyphrases: Intrusion Detection System, Metaheuristic, deep learning

@booklet{EasyChair:3452,
  author    = {Yi-Lin Chen and Huan Chen and Chun-Wei Tsai},
  title     = {An Effective Deep Learning Algorithm for Intrusion Detection},
  howpublished = {EasyChair Preprint 3452},
  year      = {EasyChair, 2020}}