Information Technology Reference
In-Depth Information
5 Conclusion
As described in the preceding section, MLP method has recognized them as a good
choice for any existing intrusion detection system. This paper provides a state-of-
the-art review of the applications of neural network to Intrusion Detection System.
Following
findings are signi
cant in the research review of IDS:
Arti
cial neural network based intrusion detection system development is an
important research trend in intrusion detection domain.
China has shown to be signi
cantly contributing (32 %) followed by USA
(16 %) and India (9 %) in terms of publication by af
liated country.
The conference paper (65 %) has recognized as the major type of research
documents followed by articles (30 %).
Lecture Notes in Computer Science has emerged as leading journal that pub-
lished 25 articles on IDS based on neural network (11 %) followed by 24 articles
in Computers and Security and 20 articles in Expert Systems with Applications
(9 %) journals.
Undoubtedly the computer science (49 %) is shown to be the major domain
publishing 509 articles followed by 273 articles in engineering (26 %) and 108
articles in mathematics (10 %).
The current research trend based on the number of articles published between the
years 2000
2013 has been shown to be increasing with R-squared value equals
0.9433 as a good
-
fit. The trend line for 2000
2014 is also shown to be increased.
-
In this research, we have proposed architecture based on Multi Layer Perceptron
neural network. The model builds the intrusion detection system learnt from the
patterns of KDD99 data set. Based on the identi
ed patterns, the architecture rec-
ognized attacks in the datasets using the back propagation neural network algo-
rithm. The proposed neural network approach resulted with higher detection rate, a
reduced amount of execution time. We continue our work in this direction in order
to build an efficient intrusion detection model. When the proposed Back propa-
gation neural network approach is compared with the other two approaches:
Recurrent and PCA neural network based on the common measures of performance
it is clearly visible as shown in Fig.
13
to outperform the performances of the other
two approaches. Further work will be undertaken to increase the performance of
the intrusion detection model and reduce the false alarm and ef
ciently handle the
identi
cation of correct anomaly dynamically.
Since the goal of this research was also to evaluate the performance of our
proposed approach by comparing with other six approaches available in literature in
terms of three measures of performance: detection rate, accuracy rate and compu-
tation time of the intrusion detection (Table
10
). The comparative research
findings
from Table
10
has revealed that the proposed approach has succeeded in achieving
increased rate of anomaly detection, reduced false alarm and at the same time
minimal execution time for the development of intrusion detection system. In KPCA
and SVM approach, the accuracy rate is 99.2 % (98.89 % accuracy in case of
Search WWH ::
Custom Search