Modelling of Intrusion Detection System Using Artificial Intelligence—Evaluation of Performance Measures - Complex System Modelling and Control Through Intelligent Soft Computations - page 307

has potential to computational adaptive, fault tolerant, maximizing speed, mini-
mizing error rates corresponding to human performance (Bezdek 1994 ).
Wu and Banzhaf ( 2010 ) commented that the popular domain of AI is different
from the CI. However there is neither full conformity on the exact nature of
computational intelligence nor there is any far and wide established vision on which
domain belong to CI: arti cial neural networks, fuzzy sets, evolutionary compu-

tation, arti cial immune systems, swarm intelligence, and soft computing. Majority

of these approaches are able to process the information using either supervised or
unsupervised learning algorithm. Supervised learning frequently constructs classi-

fiers known as a function mapping data observations to matching class labels for
misuse discovery from class-labeled training datasets. Classi ers are basically

viewed. On the other hand, unsupervised learning is different from supervised
learning due to non-availability of class-labeled data during the training stage and it
works on based on similarities of data points. Therefore it becomes a more suitable
approach to deal with anomaly detection.
Arti cantly in the

development of Intrusion Detection System (IDS) for anomaly detection, data
reduction from the research community. Due to large trend of internet usage in the
last decade in a more complex and un-trusted global internet environment, the
information systems are inescapably uncovered to the growing threats. Intrusion
Detection System is an approach use to respond to such threats. Diverse IDS
techniques have been proposed, which identify and alarm for such threats or
attacks. The Intrusion Detection System (IDS) generates huge amounts of alerts that
are mostly false positives. The abundance of false positive alerts makes it dif

cial Intelligence (AI) has recently been attracted signi

cult
for the security analyst to identify successful attacks and to take remedial actions.
Many of arti

cation, but they
alone are incapable of dealing with new types of attack which are evolving due to
the advent of real time data. To address with these new problems of networks,
arti

cial intelligence approach have been used for classi

cial
intelligence offers a vast range of techniques to classify these attacks. So to assist in
categorizing the degree of the threat, different arti

cial intelligence based IDS are opening new research avenues. Arti

cial intelligence techniques are
used to classify the alerts, our research work will be based on analyzing the existing
techniques and in the process identifying the best algorithm for the development of
an ef cient intrusion detection system.

The fundamental objectives of our contribution will be to explore for an optimal
intrusion detection system model based on Arti

cial Intelligence techniques and
evaluation perspective for performance of such predictive classi cation system.

Therefore, the objective is basically to provide solutions in developing a complex
system model. The principal chapter objectives of this research work can be
summarized as:

1. Undertake detailed study on anomaly based intrusion detection systems.
2. Exploring the research trend for security challenges of ID based on anomaly
detection after critical appraisal of the existing methodologies for intrusion
detection system.