Biomedical Engineering Reference
In-Depth Information
The above graph represents the assessment result of Data set 1as KNN and KNN-
DS method as included parameters, i.e, Accuracy, Precision, Recall.
DATA SET 2
The above graph represents the assessment result of Data set 2as KNN and KNN-
DS method as included parameters, i.e., Accuracy, Precision, Recall.
Conclusion
Our dissertation presents the performance of the intrusion detection system on
application of our new design technique. We have designed an intrusion detection
system using KNN classification and Dempster theory for detecting intrusion
behavior within the network. As compared to the earlier technique used, the com-
bined use of KNN and Dempster theory, it is found that, the performance gets
considerably enhanced. This improvised design technique gives more efficient
results. It was observe that KNN and Dempster can perform better in almost all
situations, which is further proven by comparing the result on KDD Data set 99. Our
Experiment on different data sets classifies the data using KNN classification
(Normal Packet, DOS, R2L, U2R, Probes) and later the factor of evidence is for-
mulated by using DS theory. The new pattern of intrusion is compared with the
existing pattern of intrusion and generates a new schema of pattern and updates a list
of pattern of intrusion detections and improves the true rate of intrusion detection.
Future Work
This work can be extended by studying the nitty-gritty of data mining techniques
and the fundamentals of intrusion detection system and network behavior patterns.
As a piece of future work, our design can be clubbed up with a more optimized
classification technique. This improvised structure will increase the efficiency and
will give improvised results; in addition the design can be made more compre-
hensive
by supervising
data
from
varied
data
sources
and
examining
more
complicated intrusion network scenarios.
Search WWH ::




Custom Search