An Optimal Auto Encoded Deep Neural Network based Intrusion Detection Systems for Mobile ADHOC Networks

Authors

  • Bosco Paul Alapatt
  • Anupama Jims
  • Felix M Philip

Abstract

Using the application of ad-hoc networks, communication models in this field of wireless networks have been developed. Greater research is performed for mobile nodes in mobile ad hoc networks (MANET). Intrusion Detection Systems (IDS) is considered as a main component of secured system. A major issue in security system is, it is assumed to be inefficient intrusion detection system due to the access of enormous network information. Traditional IDS provides lower detection rate as well as greater negative alarms with maximum processing time. This study provides an effective IDS method for MANET by combining feature selection (FS) based classifier approach model for efficient identification of intruders. For FS, particle swarm optimization (PSO) algorithm is utilized to pick the essential features from accessible ones. The minimized feature has the subset as and is fed to Auto encoded Deep Neural Network (AEDNN) for discovering the availability of hackers. By including PSO before classification process, it would improve the effectiveness of AEDNN. For practical experience, KDD'99 database is deployed in order to validate the projected technique. The end results signify that greater outcome of PSO-AEDNN model is attained across previous IDS in various estimating variables.

Downloads

Published

2020-03-27

Issue

Section

Articles