Machine Learning with Applications (Dec 2021)

Genetic Algorithm based feature selection and Naïve Bayes for anomaly detection in fog computing environment

  • John Oche Onah,
  • Shafi’i Muhammad Abdulhamid,
  • Mohammed Abdullahi,
  • Ibrahim Hayatu Hassan,
  • Abdullah Al-Ghusham

Journal volume & issue
Vol. 6
p. 100156

Abstract

Read online

The sharp rise in network attacks has been a major source of concern in cyber security, particularly that now internet usage and connectivity are in high demand. As a complement to cloud computing, fog computing can offer low-latency services among users of mobile and the cloud. Because of the closeness of the end users to the fog nodes and having inadequate computing resources, fog devices may get into security issues. Conventional network threats may demolish the fog computing system. The use of Intrusion Detection Systems (IDS) in conventional networks has been extensively researched, applying them directly in to the fog computing platform might become unsuitable. Nodes of the fog generate enormous quantities of data most of the time, so implementing an Intrusion detection system model over large datasets in the fog computing setting is critical. To combat some of these network attacks, an intrusion detection system (IDS), a strategic intrusion prevention innovation that can be applied in the fog computing platform utilizing machine learning techniques for network anomaly detection and network event classification threat, has proven efficient and effective. This paper presented a Genetic Algorithm Wrapper-Based feature selection and Nave Bayes for Anomaly Detection Model (GANBADM) in a Fog Environment which removes extraneous attributes to reduce time complexity while also developing an enhanced model that can predict results with greater accuracy using the Security Laboratory Knowledge Discovery Dataset (NSL-KDD). Based on the analysis, the developed system has a higher overall performance of 99.73% accuracy, with a false positive rate as low as 0.6%. This results show that the proposed GANBADM approach performs better than similar approaches in the literature.

Keywords