Modelling of Intrusion Detection System Using Artificial Intelligence—Evaluation of Performance Measures - Complex System Modelling and Control Through Intelligent Soft Computations - page 312

Recurrent Neural Network out-performs Feed-forward Neural Network, and
Elman Network for detecting attacks in a communication network (Anyanwu et al.
2011 ).
Theory and experiment show that Radial basis function network (RBFN)
algorithm has better ability in intrusion detection, and can be used to improve the
ef ciency of intrusion detection, and reduce the false alarm rate (Peng et al. 2014 ).

Binary Genetic Algorithm (BGA) as a feature extractor provide input for the
classi

cation task to a standard Multi-layer Perceptron (MLP) classi er that resulted

with very high classi cation accuracy and low false positive rate with the lowest

CPU time (Behjat et al. 2014 ).
Using k-means clustering, Naive Bayes feature selection and C4.5 decision tree
classi cation for pinpointing cyber attacks resulted with a high degree of accuracy

(Louvieris et al. 2013 ). Comparing the traditional BP networks and the IPSO-
BPNN algorithm to simulate results of the KDD99 CUP data set with the intrusion
detection system has demonstrated the BPN resulted with less time, better recog-
nition rate and detection rate (Zhao et al. 2013 ).
Feizollah et al. ( 2014 ) evaluated ers, namely Naive

Bayes, k-nearest neighbour, decision tree, multi-layer perceptron, and support
vector machine in wireless sensor network (WSN). A critical study has been made
using genetic algorithm, arti

five machine learning classi

cial immune, and arti cial neural network (ANN)

based IDSs approaches (Yang et al. 2013 ).
A network IDS applied discretization with genetic algorithm (GA) as a feature
selection to assess it ' s performance several classi ers algorithms like rules based

classi ers (Ridor, Decision table), trees classi ers (REPTree, C 4.5, Random

Forest) and Na er have been used on the NSL-KDD dataset (Aziz

et al. 2012 ; Eid et al. 2013 ). Research revealed that discretization has a positive
impact on the time to classify the test instances and is found to be an important
factor for developing a real time network IDS.
Therefore only a detail analytical view on applications of neural network based
intrusion detection system can quantitatively enlighten on the trend of kind of
diverse research based on neural network as explored in literature.

ï ve bays classi

2.1 Analysis of IDS Research Based on the Neural Network
Algorithm

This paper provides a state-of-the-art review of the applications of neural network
to IDS. The following query string has been searched using scopus search engine:
(TITLE-ABS-KEY (intrusion detection system) AND SUBJAREA (mult OR ceng
OR CHEM OR comp OR eart OR ener OR engi OR envi OR mate OR math OR
phys) AND PUBYEAR > 1999) AND (neural network). It resulted with 2,185
articles and only the relevant information has been collected to interpret the sig-
ni

cance of IDS research using neural network during the period 2000 - 2014.

Figures 3 , 4 , 5 , 6 and 7 organizes this review of the literature.