scispace - formally typeset

Journal ArticleDOI

Enhancing MLP Performance in Intrusion Detection using Optimal Feature Subset Selection based on Genetic Principal Components

01 Mar 2014-Applied Mathematics & Information Sciences-Vol. 8, Iss: 2, pp 639-649

AboutThis article is published in Applied Mathematics & Information Sciences.The article was published on 2014-03-01 and is currently open access. It has received 3 citation(s) till now. The article focuses on the topic(s): Feature (computer vision) & Selection (genetic algorithm).

Summary (2 min read)

1 Introduction

  • Security breach in network today has become one of the major problems since a single of it may cause a significant loss or damage to the information systems.
  • A number of previous intrusion detection techniques, in response, have attempted to focus on the issues of feature extraction and classification, yet less importance unfortunately has been given to the serious issue of feature selection.
  • In past, a subset of features was selected using PCA based on some percentage of the top principal components.
  • The standard dataset, KDD cup dataset, also was used to validate the proposed model.
  • The experimental results then illustrate significant performance improvements.

3 Proposed Model

  • The proposed model consists of four sections; Dataset, Feature selection, Classification and Training & Testing.
  • The selection of dataset is considered as an important issue in the intrusion detection considering that an accurate dataset can result in accurate results.
  • Dataset collection could be conducted in several ways; those are real time, simulated and test-bed, each of which has various issues.
  • Feature selection, meanwhile, was accomplished through GA and PCA due to their proven ability in feature selection.
  • The selected features sets were presented to the classifier to determine their sensitivity and importance.

3.1 Dataset

  • This work used KDD cup dataset, considered as a standard in the evaluation of intrusion detection techniques.
  • From the dataset, 20.0 0 connections were randomly selected in which each connection of raw dataset consisted of 41 features.
  • Thus, the features remained 38 in each record.

3.2 Feature selection

  • The suitable feature set simplified the classifier architecture as well as improved its overall performance.
  • The feature selection process and its applied techniques; PCA and GA are explained respectively as follows.
  • Assuming a population of size N, the offspring then doubled the size of the population and selected the best top 10 percent individuals from the combined parent-offspring population.
  • For one-point crossover, the parent chromosomes were divided at a common point chosen randomly and the resulting sub-chromosomes are swapped.
  • Therefore, the fitness evaluation contains two terms:(i) accuracy and(ii) the number of selected features.

3.3 Classification

  • The selected features were presented to MLP for classification.
  • First, its processing elements (PEs) or neurons are nonlinear.
  • Second, they are massively interconnected such that any element of a given layer feeds all the elements of the next layer.
  • MLP architecture used consists of three layers; namely input, hidden and output.
  • Algorithm Input: training−examples,η ,φ ,net Output:trainednetwork Initialize all weights o f net; for each pair<−→x ,−→t >∈ training-examples do Step 1:Forward phase:.

3.4 Training and Testing

  • In the training phase, input patterns and desired outputs are given related to each input vector.
  • To achieve this goal, weights are updated by carrying out certain steps known as training.
  • Testing of trained system involves two steps;(i) verification step, and(ii) generalization step.
  • The parametric specification used for MLP architecture during testing phase is given in Table2.
  • Thus, research work achieved this objective by using GA and PCA that made the classifier simpler as well as more efficient in performance.

4 Experimental Results

  • The MLP based intrusion analysis engine was evaluated on different feature subsets.
  • This section presents MLP results and their sensitivity analysis in different scenarios.
  • First of all, MLP was tested on original dataset without c© 2014 NSP Natural Sciences Publishing Cor. using PCA and GA, which consisted of 38 features.
  • Five thousand exemplars or input samples were randomly selected from twenty thousand dataset.
  • Five thousand exemplars contained two types of connections; normal and intrusive, in which 3,223 were normal and 1,777 were intrusive.

4.1 Comparison with existing Approaches

  • Comparison of the performance of the developed system is done with some other intrusion detection approaches introduced in related work section.
  • The SVM converges to the optimal solution in 1000 epochs while MLP converge in 173 epochs.
  • Therefore, MLP is considered a good classifier for intrusion analysis due to its proven ability to handle large data such as traffic data on networks, less number of epochs and time in training process.
  • This process reduced the number of features to ten features as compared to previous approach [13] which has twelve features.
  • Table 7 shows the comparative analysis of applied approach with other approaches.

5 Conclusion

  • A performance enhancement model is proposed for intrusion detection system based on an optimal feature subset selection using several genetic c© 2014 NSP Natural Sciences Publishing Cor. principal components.
  • The feature selection has been accomplished using the techniques of PCA and GA.
  • The selected principal components called genetic principal components are the basis of feature subsets.
  • The KDD-cup dataset used is a benchmark to evaluate the security detection mechanisms.
  • The performance of applied approach was then addressed.

Did you find this useful? Give us your feedback

...read more

Content maybe subject to copyright    Report

Appl. Math. Inf. Sci. 8, No. 2, 639-649 (2014) 639
Applied Mathematics & Information Sciences
An International Journal
http://dx.doi.org/10.12785/amis/080222
Enhancing MLP Performance in Intrusion Detection
using Optimal Feature Subset Selection based on
Genetic Principal Components
Iftikhar Ahmad
Department of Software Engineering, College of Computer and Information Sciences, King Saud University, P.O. Box 51178, Riyadh
11543, Saudi Arabia
Received: 26 Mar. 2013, Revised: 27 Jul. 2013, Accepted: 29 Jul. 2013
Published online: 1 Mar. 2014
Abstract: Computer and network systems nowadays are facing many security issues, one of which considered important is intrusion.
To prevent such intrusion, a mechanism for optimal intrusion detection is deemed necessary. A number of tools and techniques are
available, yet most of them still face a main problem that is on performance. The performance, in essence, can be increased by reducing
false positives and increasing accurate detection rate. What has made the performance terrible in the existing intrusion detection
approaches is due to the use of a raw dataset that includes redundancy and leads the classifier to be confused. To overcome this
issue, Principal Component Analysis (PCA) has been used to project a number of raw features on principal feature space and to
select the features based on their sensitivity determined by the magnitude of eigenvalues. Here, only the features corresponding to the
highest eigenvalues are selected; the remaining features, by contrast, are ignored. Due to the ignorance of many important and sensitive
features for the classifier for their lowest eigenvalues, this method comes to be not optimal. Therefore, a suitable method is necessary
to select a subset of features, which, in turn, can enhance the classifier performance. The focus of this research is to observe a space of
principal features to find a subset of sensitive features to the classifier, which can optimize the detection accuracy. Genetic Algorithm
(GA) has been applied to solve an optimization problem. The raw features have, afterwards, been transformed through PCA into
principal features space. GA, in this case, was used to search this features space to obtain principal components called genetic principal
components (GPC). The feature set obtained through this process was, in turn, presented to the classifier. The Multilayer Perceptron
(MLP), meanwhile, was used for classification considering its proven ability. Additionally, Knowledge Discovery and Data mining
(KDD) cup dataset was used for the validation of the proposed approach, which is considered as a benchmark to evaluate the intrusion
detection approaches. The performance of this approach has been analyzed and compared with a number of existing approaches. The
results then show that proposed method outperforms the existing approaches. Not only does it significantly reduce the dimension of the
feature space but also improves the detection accuracy.
Keywords: Intrusion Detection System (IDS), Multilayer Perceptron (MLP), Principal Component Analysis (PCA), Genetic
Algorithm (GA), Genetic Principal Component (GPC), Detection Rate (DR), False Positive (FP), False Negative (FN), True Negative
(TN), True Positive (TP) and Dataset.
1 Introduction
Security breach in network today has become one of the
major problems since a single of it may cause a
significant loss or damage to the information systems.
Hence, an effective intrusion detection system is deemed
essential in use to address such incidents or intrusions. In
response, a wide variety of intrusion detection systems
are now available. However, the main issue on their poor
performance still remains due to false positives.
One of the reasons for this is related to the use of raw
dataset. The performance of intrusion detection
principally can be increased by using a proper feature
selection method. A number of previous intrusion
detection techniques, in response, have attempted to focus
on the issues of feature extraction and classification, yet
less importance unfortunately has been given to the
serious issue of feature selection. In past, a subset of
features was selected using PCA based on some
percentage of the top principal components. Further, the
Corresponding author e-mail:
wattoohu@gmail.com
c
2014 NSP
Natural Sciences Publishing Cor.

640 I. Ahmad: Enhancing MLP Performance in Intrusion Detection...
features corresponding to the highest eigenvalues were
selected and those corresponding to the lowest ones, by
contrast, were ignored. This method in selecting an
appropriate set of features might be not effective for being
potential to omit certain features that, for their sensitivity,
might be very important to the classifier. To cope with this
problem, GA is proposed for selecting good subsets of
features from the PCA space.
This study clarifies that feature selection is a
significant issue in intrusion detection. Further, the
combination of PCA and GA provide a simple, general,
and powerful framework in selecting a number of
important and sensitive features, leading to improved
performance of the classifier. In this work, raw features
have been presented to PCA for transformation to
principal feature space. This transformation makes such
features more visible and organized in PCA feature space.
At this point, the feature space was searched through GA
to find some optimal features based on genetic
eigenvectors. The resultant feature set was then presented
to the classifier. On the other hand, MLP (Multilayer
Perceptron) was used as a classifier and tested on the
selected feature set. The standard dataset, KDD cup
dataset, also was used to validate the proposed model.
The experimental results then illustrate significant
performance improvements.
The rest of the paper is organized as follows. Section
2 is designed to present some related work. It is then
followed by Section 3 dealing with the proposed model
including the explanation of dataset, feature selection
process using PCA and GA, classification and training
and testing. Section 4 and Section 5 are to discuss the
experiments and results and to draw a conclusion
respectively.
2 Related Work
Feature selection is a serious issue in intrusion detection.
In some previous approaches, it was considered
insignificant and depended on powerful classification
algorithms to deal with redundant and irrelevant features.
Moreover, feature extraction and classification have been
more focused in intrusion detection in which the features
have been extracted from the raw features using PCA.
The raw features were then projected to principal space to
select a subset of features. Features corresponding to
highest eigenvalues were included in the subset, and those
corresponding to the lowest eigenvalues, oppositely, were
ignored. In this method, some percentage of the total
features was included in the feature subset [
11,12,13].
In fact, this is not an effective scheme to select an
appropriate set of features in this space in consideration to
the potential of missing important and sensitive features
to the classifier. Related work in intrusion detection is
presented in which the prime concentration is on
classification.
In [
2], PCA and neural networks have been used to
detect intrusion. The PCA here was applied for
classification and neural networks for online computing.
The features were selected based on 22 principal
components while others were ignored for being less
important - the importance of features was determined
based on the highest eigenvalues. Such feature selection,
for some reasons, is not effective for a possibility to miss
many important features having sensitive information for
classifier or intrusive analysis engine [
11,12].
The importance of a feature is defined differently in
the existing research work of intrusion detection. It can be
determined based on eigenvalues, accuracy, detection
rate, and false alarms. In [3], the importance of a feature
is determined based on the accuracy and the number of
false positives of the system with and without the feature.
The feature selection was based on method:
leaving-one-out; removing one feature from the feature
set, performing experiment, and comparing the new
results with the original result. If any case of these
described cases occurs, the feature is regarded as
important; otherwise, unimportant. To illustrate, with 200
features in the original set, an experiment might be
repeated 200 times to ensure whether each feature is
important or not. As a consequence, not only does this
method involve complexity but also overheads on huge
dataset.
In [
4], real-time pattern classification has been
performed using a radial basis function (RBF) network.
The Elman network here was used to restore the memory
of past events and full featured DARPA dataset was used
in experiment of this work. Such method, consequently,
might increase training and testing overheads on the
system as well as make the classifier confused to produce
false alarms.
In [
5], PCA was used to determine a subset of features
based on a feature reduction concept. The feature
reduction accelerated training and testing process for the
classifier. However, this can affect on the efficiency and
accuracy of the system. For example, few numbers of
principal components speed up the training efficiency
while a large number of principal components make the
classifier confused to produce false alarms. Such
compromise in fact is not suitable in intrusion detection
mechanism. Hence, other method suitable in feature
section to avoid such compromise is deemed essential.
In [
6], GA was in use for features and parameters
selection for intrusion detection model. The Support
Vector Machines (SVM) has been applied as an intrusive
analysis engine. Even though this method was capable of
minimizing a number of features and maximizing the
detection rates,, features uniformity still become the
problem. Since the features in original forms are not
consistent, they must be transformed into a feature space
for a well organized form.
In [
11], a mechanism of intrusion detection was
proposed using a number of soft computing techniques:
SVM, GA and PCA. The proposed model was tested on
c
2014 NSP
Natural Sciences Publishing Cor.

Appl. Math. Inf. Sci. 8, No. 2, 639-649 (2014) / www.naturalspublishing.com/Journals.asp 641
two subsets of features, one of which was obtained using
PCA and GA while another one was selected on the basis
of top percentage of principal components using PCA
only. The first subset consisted of 12 features while the
second one did consist of 22 features all directly taken
from the principal space. Both sets were then tested on
SVM, and compared for their performance. The focus
was performance comparison on both feature sets. In fact,
this approach was still needed further experimentation to
validate the proposed model.
The above mentioned approach was explored further
in [
15]. The principal space was concentrated on selecting
features for a suitable subset. The GA was used to search
the PCA space for feature selection. Several experiments
were conducted for optimal feature subset selection, and
their results were analyzed. The results showed that the
proposed approach outperformed the existing approaches.
This work used SVM which is not suitable for large dataset
and it also increases training time of the classifier.
In [
13], an attempt of a feature subset was initiated
using the MLP as classifier. However, the presented
approach was not sufficient to verify the concept. Three
different subsets were selected with 12, 20 and 27
features. These subsets were tested on MLP, and their
results were analyzed. This approach had a number of
issues, which were not considered and tested to validate
the concept. For example, the classifier may perform
better on the feature sets consisting of raw features,
transformed features and conventionally PCA selected
features. This work then will be further explored,
extended and verified in the model proposed in this
research.
3 Proposed Model
The proposed model consists of four sections; Dataset,
Feature selection, Classification and Training & Testing.
The selection of dataset is considered as an important
issue in the intrusion detection considering that an
accurate dataset can result in accurate results. Dataset
collection could be conducted in several ways; those are
real time, simulated and test-bed, each of which has
various issues. To avoid such issues, the standard dataset;
KDD cup dataset has been used to validate the model of
this work. Feature selection, meanwhile, was
accomplished through GA and PCA due to their proven
ability in feature selection. The selected features sets were
presented to the classifier to determine their sensitivity
and importance. Further, the classification was performed
through a well-known classifier such as MLP. The
classifier, subsequently, was trained and tested to analyze
the feature subsets. Figure 1 illustrates the proposed
model. The detail of the proposed model is described as
follows.
Fig. 1: Proposed model for intrusion detection
3.1 Dataset
This work used KDD cup dataset, considered as a
standard in the evaluation of intrusion detection
techniques. From the dataset, 20.000 connections were
randomly selected in which each connection of raw
dataset consisted of 41 features.
f
1
, f
2
, f
3
, f
4
, .. f
n
where n = 41 (1)
Three symbolic features; protocol-type, service and flag
were discarded for having no any impact on the classifier.
Thus, the features remained 38 in each record.
f
1
, f
2
, f
3
, f
4
, .. f
m
where m = 38 (2)
3.2 Feature selection
The feature selection is an important task in this work.
The suitable feature set simplified the classifier
architecture as well as improved its overall performance.
Figure
2 presents the flow of feature selection algorithm.
The feature selection algorithm used two types of
techniques: PCA and GA, which has been being widely
used in the process of feature selection in many various
fields such as image processing, data mining or medical.
Below is the algorithm proposed for feature selection.
Algorithm
Step 1: Let Fs is a feature set, which consists of 38
features.
Fs = f
1
, f
2
, f
3
, f
4
, f
m
where m = 38 representing a total
c
2014 NSP
Natural Sciences Publishing Cor.

642 I. Ahmad: Enhancing MLP Performance in Intrusion Detection...
Fig. 2: Feature subset setection algorithm flow
number of features.
Step 2: The feature set (Fs) is projected into another
feature space called the principal feature space. This can
be expressed as
Fs PCA
N=38
i=1
PCs
i
Step 3: The principal feature space is searched through
GA for the selection of genetic eigenvectors. This is
represented in the following expression.
PCs GA
r
k=1
GPCs
k
Step 4: A set of features (F
s) is obtained based on
genetic principal components, which are the subset of
original feature set (Fs). This can be expressed as.
GPCs
k
F
s
l
F
s
l
Fs
The term r represents random number and l indicates the
number of features less than 38. This step is repeated
several times until obtaining a set with a maximum
accuracy and minimum number of features.
The feature selection process and its applied
techniques; PCA and GA are explained respectively as
follows. PCA is a valuable statistical method that has an
application in many fields such as face recognition and
image compression. It is a common technique in finding
patterns in high dimension data. It has also been used to
analyze a large data set and the relationship between the
individual points in that set [
4,11,22]. PCA purposely is
to reduce the dimension of the data while retaining as
much as possible of the variation present in the original
dataset. It provides a way of identifying patterns in data,
and expressing the data in such a way as to highlight their
similarities and differences [
6,11]. However, PCA here
was used to transform the input vectors to the new search
space. On the other hand, the choosing of number of
principal components is done by GA. The flow of applied
PCA is sown in Figure
3 . Below is the PCA algorithm
applied in this work.
Algorithm
Let x
1
, x
1
, x
1
, x
1
, ..x
M
are N × 1 vectors
Step 1: ¯x =
1
M
M
i=1
x
i
Fig. 3: PCA algorithm flow
Step 2: Subtract the mean:
φ
(x1 ¯x)
Step 3: From the matrix A = [
φ
1,
φ
2,
φ
3.....
φ
M] (N M)
Matrix then compute C =
1
M
M
N=1
φ
n
φ
n = AA
T
Step 4: Compute the eigenvalues of
C :
λ
1 >
λ
2 > ......
λ
N
Step 5: Compute the eigenvectors of
C =
µ
1
,
µ
2
......
µ
N
Since C is symmetric, C =
µ
1
,
µ
2
......
µ
N
form a basis i.e.
any vector x or actually (x1 ¯x) can be written as a linear
combination of the eigenvectors.
(x
1
¯x) = b
1
µ
1
+ b
2
µ
2
+ ..... +b
N
µ
N
=
N
i=1
b
i
µ
i
In order to overcome the issue in optimal feature
selection, GA was applied to search the principal
components space in order to select an optimal subset of
features. This is a main contribution that positively impact
on the performance of intrusion detection analysis engine.
GA is inspired by the biological mechanisms of
reproduction [
5,11,12,13,24]. GAs operate iteratively on
a population of structures, each of which represents a
candidate solution to the problem.
In this work, the initial population was generated
randomly in which each individual approximately
contained the same number of 1s and 0s. All experiments
used a population size of five thousands and hundred
generations. In most cases, the GA converged in less than
hundred generations [
11,13]. The GA started its search
on the initial population. Three basic genetic operators,
namely selection, crossover, and mutation have guided
this search. The genetic search process is iterative:
evaluating, selecting, and recombining strings in the
population during each iteration or generation until
reaching some termination conditions. The applied GA
algorithm is given.
Algorithm
Step 1: Create initial population. The chromosomes are
c
2014 NSP
Natural Sciences Publishing Cor.

Appl. Math. Inf. Sci. 8, No. 2, 639-649 (2014) / www.naturalspublishing.com/Journals.asp 643
Fig. 4: GA algorithm flow
selected randomly from principal components.
Step 2: Evaluate the population
Step 3: Check the termination condition?
i. If condition is satisfied than select best individuals
ii. else go to step 2 (selection crossover mutation)
Step 4: Find subset of genetic principal components
Step 5: Select feature subset
Each string was evaluated based on a fitness function
determining the suitability of the candidate solutions.
This was inspired by the idea of survival of the fittest in
natural selection. Selecting a string depends on its fitness
relative to other strings in the population. So, strings with
high fitness were selected and the remaining ones with
low fitness were removed from the population. Selection
acted as a filter, removing poor performance solutions and
selected high performance one. Further, selection as a
genetic operator chose chromosomes from the current
generations population for inclusion in the next
generations population. Five selection operators; namely
roulette, tournament, top percent, best and random were
used. Tournament used roulette selection N times (the
Tournament Size) to produce a tournament subset of
chromosomes. The best chromosome in this subset was
then chosen as the selected chromosome. This selection
method applied an additional selective pressure over plain
roulette selection. There was also an option to specify
whether the chance of being selected was based on fitness
or on rank. In Best, the best chromosome was selected (as
determined by the lowest cost of the training run). If there
are two or more chromosomes with the same best cost,
one of them is chosen randomly. In random, a
chromosome from the population was randomly selected.
Similarly, in top percent, a chromosome from the top N
percent (the Percentage) of the population was randomly
selected. Top percent selection method was used in the
experiments for giving a better performance compared to
other selection operators. So, the selection strategy was
GA generational. Assuming a population of size N, the
offspring then doubled the size of the population and
selected the best top 10 percent individuals from the
combined parent-offspring population.
Before making into the next generations population,
selected chromosomes may undergo crossover and
mutation. Fundamentally, crossover is categorized into
three: one-point crossover, two-point crossover, and
uniform crossover [
12]. For one-point crossover, the
parent chromosomes were divided at a common point
chosen randomly and the resulting sub-chromosomes are
swapped. For two-point crossover, the chromosomes were
thought of as rings with the first and last gene connected.
In this case, the rings were divided into two common
points chosen randomly and the resulting sub-rings were
swapped. Uniform crossover, meanwhile, was different
from the above two schemes. In this case, each gene of
the offspring was selected randomly from the
corresponding genes of the parents. For simplicity, we
used one-point crossover here. The crossover probability
used in all experiments was 0.9.
Crossover is applied with high probability and allows
information exchange between points. Its goal is to
preserve the fittest individuals without introducing any
new value. Mutation, in contrast, is a low probability
operator, which flips a specific bit to restore the lost
genetic material [
11,12].
Mutation is a genetic operator that alters one or more
gene values in a chromosome from its initial state [13].
This can result in entirely new gene values being added to
the gene pool. With these new gene values, the genetic
algorithm may be able to arrive at a better solution than
the previous one. Mutation is an important part of the
genetic search as it helps to prevent the population from
being stagnant at any local optima. It occurs during
evolution according to the defined probability. This
probability should usually be set fairly low. If it is set too
high, the search will turn into a primitive random search
[
3]. The traditional mutation operator is used which just
flips a specific bit with a very low probability. The
mutation probability used in all experiments was 0.01.
Crossover and mutation generate new solutions for
exploration through string operations. Genetic algorithms
do not guarantee a global optimum solution. However,
they have an ability to search through very large search
spaces and come to nearly optimal solutions fast [
1].
c
2014 NSP
Natural Sciences Publishing Cor.

Citations
More filters

Journal ArticleDOI
TL;DR: A feature subset selection based on PSO is proposed which provides better performance as compared to genetic algorithm, which has been used to search the most discriminative subset of transformed features.
Abstract: The prevention of intrusion in networks is decisive and an intrusion detection system is extremely desirable with potent intrusion detection mechanism. Excessive work is done on intrusion detection systems but still these are not powerful due to high number of false alarms. One of the leading causes of false alarms is due to the usage of a raw dataset that contains redundancy. To resolve this issue, feature selection is necessary which can improve intrusion detection performance. Latterly, principal component analysis (PCA) has been used for feature reduction and subset selection in which features are primarily projected into a principal space and then features are elected based on their eigenvalues, but the features with the highest eigenvalues may not have the guaranty to provide optimal sensitivity for the classifier. To avoid this problem, an optimization method is required. Evolutionary optimization approach like genetic algorithm (GA) has been used to search the most discriminative subset of transformed features. The particle swarm optimization (PSO) is another optimization approach based on the behavioral study of animals/birds. Therefore, in this paper a feature subset selection based on PSO is proposed which provides better performance as compared to GA.

42 citations


Cites background from "Enhancing MLP Performance in Intrus..."

  • ...But there is possibility to miss several important features and to include irrelevant features in feature subset during this process [3]....

    [...]


Book ChapterDOI
01 Jan 2014
TL;DR: Four methods for making rational decisions by either marginalizing irrelevant information or not using irrelevant information are discussed, which are marginalization of irrationality approach, automatic relevance determination, principal component analysis and independent component analysis.
Abstract: This chapter deals with the concept of using relevant information as a basis of rational decision making. In this regard, whenever information is irrelevant it needs to be marginalized or eliminated. Making decisions using information which contains irrelevant information often confuses a decision making process. In this chapter we discuss four methods for making rational decisions by either marginalizing irrelevant information or not using irrelevant information. These methods are marginalization of irrationality approach, automatic relevance determination, principal component analysis and independent component analysis. These techniques are applied to condition monitoring, credit scoring, interstate conflict and face recognition.

1 citations


Journal Article
TL;DR: A simple and fast face recognition system on the basis of feature extraction using Principal Component Analysis (PCA) is proposed, a classical and successful method of dimension reduction and the discrete cosine transform paper (DCT) is a well know compression technique.
Abstract: Face recognition is one of the most successful applications of image analysis and understanding and has gained much attention in recent years. Face Recognition is a computer based application which detects the faces of different persons for authentication and various other purposes. It has separate applications from the fingerprint and iris recognitions. There are various successful techniques are purposed so far as Holistically methods and Discrete Cosine Transform (DCT). In this paper, we purpose a simple and fast face recognition system on the basis of feature extraction using Principal Component Analysis (PCA) is a classical and successful method of dimension reduction and the discrete cosine transform paper (DCT) is a well know compression technique. This paper proposes for improving the recognition rate of the face recognition system. The main advantage of this technique is increase the efficiency and implementation easier, high speed and better recognition rate .

Frequently Asked Questions (1)
Q1. What are the contributions mentioned in the paper "Enhancing mlp performance in intrusion detection using optimal feature subset selection based on genetic principal components" ?

To overcome this issue, Principal Component Analysis ( PCA ) has been used to project a number of raw features on principal feature space and to select the features based on their sensitivity determined by the magnitude of eigenvalues. The focus of this research is to observe a space of principal features to find a subset of sensitive features to the classifier, which can optimize the detection accuracy.