Deep Learning-Based AI Modeling, Intrusion Detection

Deep Learning-Based AI Modeling, Intrusion Detection

Madhab Paul Choudhury, Madhab Paul Choudhury, Chandrashekhar Azad
Copyright: © 2022 |Pages: 16
DOI: 10.4018/978-1-6684-4558-7.ch005
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Machine learning techniques are being used to create an intrusion detection system (IDS) for detecting and classifying cyber-attacks at the network-level and the host-level in a timely manner. Various datasets are available for research by cyber security researchers. However, no previous study has shown the detailed analysis of the performance of various machine learning algorithms on various available datasets. As the nature of malware is changing dynamically with the changing attacking methods, the detailed analysis of the available data sets is necessary to find out the cause of the malware datasets, and accordingly, necessary steps can be executed for maintaining the security of the network. A deep neural network (DNN) is being explored to develop an effective intrusion detection system. The optimal network parameters and network topologies for DNNs are chosen through the following hyper parameter selection methods with KDD Cup 99 dataset. The DNN model can be applied on KDD Cup 99 and on other datasets also such as NSL-KDD, UNSW-NB15, Kyoto to conduct the experiment.
Chapter Preview
Top

The research activity on the intrusion detection system has been discussed as follows: (Vinayakumar et al., 2019) have proposed the holistic approach to obtain a real time intrusion detection system using the deep learning technique. Here the information regarding the current network traffic attacks have been furnished. The authors have tried to scale the performance of a Network anomaly detection using KDD 99 and varied KDD cup 99 data set. With the KDD cup 99 datasets none of the machine learning classifiers are capable to improve the attack detection rate. DARPA/KDD Cup 88 data set have been failed to evaluate the classical intrusion detection system (IDS). But KDD Cup 99 is the most widely used reliable benchmark dataset in most of the study related to intrusion detection (ID) system evaluation and related other security related tasks. To resolve the inherent issues that exist in the KDD cup 99, one refined version NSL KDD has been used. Still this refined version is not able to solve the entire problem, so some more extra features have been added to KDD cup 99. This new dataset is called KYOTO dataset. The Kyoto datasets do not contain the false positives that minimize the number of alerts to the network administration. Most widely used dataset for HIDS is KDD Cup 98, KDD Cup 99.

(Moradi et al., 2004) have proposed a neural network-based system for Intrusion Detection and Classification of attacks to solve a multi class. For that reason, an output layer with three states have been used. Here three layers of neural network have been used. Under Matlab neural network toolbox has been used for MLP networks.

Complete Chapter List

Search this Book:
Reset