Modelling of Intrusion Detection System Using Artificial Intelligence—Evaluation of Performance Measures - Complex System Modelling and Control Through Intelligent Soft Computations - page 326

TSS

Kappa Statistic

Overall Accuracy

Specificity

Sensitivity

0.8400 0.8600 0.8800 0.9000 0.9200 0.9400 0.9600 0.9800 1.0000

Value of measure

BPN RECURRENT PCA

Fig. 13 Comparison of the measure of performance for three neural network architecture applied
to develop intrusion detection system

Table 10 Comparative performance of literature available approaches used with proposed
multilayer perceptron approach based on detection rate, accuracy and computation time
Approach used References Detection

rate test-
ing %

Accuracy
testing %

Computation
time (s)

Dataset used in
experiment

KPCA and SVM Kuanf et al.

( 2012 )

- 99.2

(training:
99.975)

407.918466 KDD dataset

6000 sample-
4000 for training,
2000 for testing
(Han 2012 )

Resilient back
propagation neural
network

Naoum et al.
(2012)

94.7 - - KDD dataset

(Naoum et al.
2005 )

Decision tree
based light weight
intrusion detection
using wrapper
approach

Sivatha Sindhu
et al. ( 2012 )

98.38 KDD dataset

(Sivatha Sin-
dhu et al. 2012 )

- -

Neural network Devaraju and

Ramakrishnan
( 2011 )

97.5 KDD dataset

(Kuanf et al.
2012 )

- -

BPNN Mukhopadhyay

et al. ( 2011 )

KDD dataset
(Mukhopadhyay
et al. 2011 )

- - -

SOM Ibrahim et al.

( 2013 )

92.37 - KDD 99

Our proposed
approach a

- 99.10 98.89 11.969 KDD 20 %

dataset

a The data is taken from 70 to 30 dataset as it is giving better detection rate