Modelling of Intrusion Detection System Using Artificial Intelligence—Evaluation of Performance Measures - Complex System Modelling and Control Through Intelligent Soft Computations - page 309

Intrusion Detection System

Detection Method Response

to intrusion

Audit Data
Source

Locus of detec-
tion System

Mi-
suse

Ano-
maly

Pas-
sive

Ac-
tive

Hos
ts

Net-
works

Distri-
buted

Cen-
tral

Fig. 1 Classification scheme of intrusion detection system taken from Wu and Banzhaf ( 2010 )

Therefore, anomaly detection has the potential of identifying latest kind of
attacks, and only necessitates normal data during generation of pro les. Though,

the main intricacy involves in determining borders among normal and abnormal
behaviors, as a result of the lack of abnormal examples during the learning stage.
An additional complexity is to familiarizing itself to continually varying normal
behavior, particularly for dynamic anomaly detection.
Additionally there are other features used to classify intrusion detection system
approach, as shown in Fig. 1 (Wu and Banzhaf 2010 ).
One frequent method applied to identify intrusion detection is by classi

cation

de er

is not to investigate the data to determine interesting partition but also to settle on
how new data will be classi

ned as dividing the samples into distinct partition. The purpose of the classi

cation grouped the
data records in a encoded classes applied as features to label each sample, dis-
criminating elements

ed. In intrusion detection, classi

fitting to anomaly or normal attack classes. However the

classi fine tuning approaches to decrease false positive

rates. Thus intrusion detection is considered as a binary categorization problem
(Liao and Vemuri 2002 ).
Artificial neural network is relatively new and emerging approach to easily deal
with complex classi

cation has to be used with

cation with much better precision and output and the con-
ceptual background of different types of arti cial neural network with diverse

application domains explored in literature are discussed in the next section.

1.2 Arti cial Neural Network

Arti cial intelligence (AI) is an interdisciplinary domain exhibits human-like

intelligence and demonstrated by hardware or software. The term AI was coined by
McCarthy et al. ( 1955 ) and de ned it as “ the science and engineering of making

intelligent machines cial Neural Network (ANN) is mas-

sively parallel interconnections of simple neurons that act as a collective system
(Haykin 2005 ). The ANNs mimic the human brain so as to perform intelligently.
The major bene

” (McCarthy 2007 ). Arti

ts include high computation rate due to their massive parallelism