X=NSL_KDD Dataset/KDD Train.txt;T= NSL_KDD Dataset/KDD Test.txt;X,T =NSL_KDD Dataset;
rng(0,’twister’); % For reproducibility
hiddenSize = 15;
autoenc1=trainAutoencoder(X,hiddenSize,’MaxEpochs’,400,..
‘L2WeightRegularization’,0.001,…
‘SparsityRegularization’,4,…
‘SparsityProportion’,0.05,…
‘DecoderTransferFunction’,’purelin’);
features1 = encode(autoenc1,X);
hiddenSize = 10;
autoenc2 = trainAutoencoder(features1,hiddenSize,…
‘L2WeightRegularization’,0.001,…
‘SparsityRegularization’,4,…
‘SparsityProportion’,0.05,…
‘DecoderTransferFunction’,’purelin’,…
‘ScaleData’,false);
features2 = encode(autoenc2,features1);
softnet =
trainSoftmaxLayer(features2,T,’LossFunction’,’crossentropy’);
deepnet = stack(autoenc1,autoenc2,softnet);
deepnet = train(deepnet,X,T);
NSL_KDD_type = deepnet(X);
plotconfusion(T,NSL_KDD_type);
”
(a)
Feature Learning
from pre-processed data
(b)
Soft-max Regression
classifier training for the derived training data
Self-taught Learning
(c)
“Classification
using self-taught learning
Figure 3
4.2. Performance Evaluation
To ascertain the performance evaluation of STL for the
NIDS, execute three types of classification
(a)
Normal and
anomaly (2-class)
(b)
Normal and four
different attack categories(5-class)
(c)
Normal and 22
different attack (23- class)
Evaluate the accuracy metric using the following
Accuracy: “which
is the percentage of correctly classified records over the total number of
records.
Precision (P): which is the percentage ratio of the number of “true
positives” (TP) records divided by the number of true positives (TP) and false
positives (FP) classified records.
P=
Recall (R): which is the percentage ratio of number of “true positives
records” divided by the number of true positives and false negatives (FN)
classified records.
R=
F-Measure
(F): The harmonic mean of precision and
recall and represents a balance between them.
F=
4.2.1 Evaluation Based on training dataset
Create a 10- fold cross validation on the training
data to classify the accuracy of STL for 2-class,5-class and 23-class.
Thereafter compare its performance with the soft-max regression when its
applied on the NSL-NDD data set without feature learning.
Accuracy for various Classification
Figure 4
If the Accuracy is evaluated
for 2, 5, and 23-classes, the STL should achieved >98% accuracy 4. for all
types
4.2.2 Evaluation Based on training with test dataset
Evaluate the STL and SMR
for class 2 and 5 – class using the test data. Perform the accuracy metric for
the STL
“Accuracy
for various Classification”
Figure 5
If applied on testing
and training data, the STL should achieved accuracy of ~88% for 2-class 4
which is far better than other previous researched methods
