X hits on this document

PDF document

Modeling Intrusion Detection Systems Using Linear Genetic Programming Approach - page 9 / 10

27 views

0 shares

0 downloads

0 comments

9 / 10

Fig. 4. Feature reduction results for probe

Fig. 5. Feature reduction results for DoS

Probe Feature Reduction Accuracies

DoS Feature Reduction Accuracies

100.5

100.2

Accuracy (%)

100

99.5

99

98.5

98

41

36

31 26 21 16 No of Features

11

6

Overall Accuracy Probe Accuracy Rest Accuracy

Accuracy (%)

100 99.8 99.6 99.4 99.2

99 98.8 98.6

41

36

31 26 21 16 No of Features

11

6

Overall Accuracy DoS Accuracy Rest Accuracy

Fig. 6. Feature reduction results for U2Su

Fig. 7. Feature reduction results for R2L

U2Su Feature Reduction Accuracies

R2L Feature Reduction Accuracies

Accuracy (%)

120 100 80 60 40 20 0

41

36

31 26 21 16 No of Features

11

6

Overall Accuracy U2Su Accuracy Rest Accuracy

Accuracy (%)

100.5

100 99.5

99 98.5

98 97.5

97

41

36

31 26 21 No of Features

16

11

6

Overall Accuracy R2L Accuracy Rest Accuracy

6 Conclusions

Table 5 summarizes the overall performance of the three soft computing paradigms considered. LGPs outperform SVMs and RBP in terms of detection accuracies with the expense of time. SVMs outperform RBP in the important respects of scalability (SVMs can train with a larger number of patterns, while would ANNs take a long time to train or fail to converge); training time and prediction accuracy. Resilient back propagation achieved the best performance among the several other neural network learning algorithms we considered in terms of accuracy (97.04 %). The performances of using the important features for each class, give comparable performance with no significant differences, to that of using all 41 features.

Table 5. Performance comparison of testing for class specific classification

Class

SVMs

RBP

LGP

Accuracy (%)

Accuracy (%)

Accuracy (%)

Normal

98.42

99.57

99.64

Probe

98.57

92.71

99.86

DoS

99.11

97.47

99.90

U2Su

64

48

64

R2L

97.33

95.02

99.47

Document info
Document views27
Page views27
Page last viewedMon Dec 05 19:23:56 UTC 2016
Pages10
Paragraphs417
Words4082

Comments