Reptile Search Algorithm Based Feature Selection Approach for
Intrusion Detection
Maher O Al-Khateeb
1
and Ali Douik
2
1
ISITCom, NOCCS-ENISO Lab, University of Sousse, Sousse, Tunisia
2
National Engineering School of Sousse, NOCCS-ENISO Lab, University of Sousse, Sousse, Tunisia
Keywords: Reptile Search Algorithm, Feature Selection, Intrusion Detection, Cybersecurity, Metaheuristic.
Abstract: In Cybersecurity, the Rise of Machine Learning (ML) Based Security Solutions Has Led to a New Era of
Defense Against Evolving Threats, with Intrusion Detection (ID) Systems at the Forefront. However, the
Effectiveness of These Systems Is Profoundly Influenced by the Quality and Relevance of the Input Features.
the Presence of Redundant Features Can Compromise Their Performance, Making Feature Selection (FS) a
Crucial Step in Optimizing ID Solutions. This Paper Uses the Reptile Search Algorithm (RSA) as a Powerful
FS Method. It Offers a Gradient-Free Approach, Avoiding Local Optima and Enabling Global Optimization.
Comparative Analysis Using Five Freely Available ID Datasets and Benchmarked Against Several Methods
Validated Superior Performance of the RSA for ID.
1 INTRODUCTION
The deployment of Internet of Things (IoT),
information technology and operational environments
have given rise to new cybersecurity risks. These risks
threaten the security of operational ecosystems, safety,
and efficiency, posing a danger to physical and
financial wellbeing (Yadav,2023). The growth of
cyber-attacks threat affects businesses, social
networks, digital privacy, and precarious
infrastructure. ID systems play a crucial role in
enhancing the security of IT infrastructures. They are
effective in detecting and countering attacks,
providing protection against intrusive hackers
(Khan,2020).
An intrusion is characterized as unexpected
activities that can harm the confidentiality, integrity,
and availability of the network. To detect anomalies,
IDs analyze network traffic and packet header fields
to identify unusual patterns, thereby preventing or
minimizing damage to the network or system
(Alsoufi,2021). The primary goal of an IDs is to
identify and avert unauthorized use and both any kind
of network intrusions, hence boosting the overall
security of the network.
IDs are typically deployed on network nodes or
hosts and use a combination of signature-based and
anomaly-based detection techniques. Signature-based
detection involves comparing network traffic or
system activity against a database of known attack
patterns or signatures (Khraisat,2019). Anomaly-
based detection involves analysing network traffic or
system activity to identify behaviours that deviate
from normal patterns. IDs can be classified into two
types: Network-based Intrusion Detection Systems
(NIDS) and Host-based Intrusion Detection Systems
(HIDS). NIDS analyse network traffic and look for
patterns that indicate an intrusion attempt, while HIDS
analyse activity on individual hosts, such as system
calls and file access patterns.
ML is a subset of artificial intelligence that
involves training algorithms to analyse and learn from
data. ML algorithms can be used to classify data, make
predictions, and identify patterns in large datasets
[(Liu,2019), (Al-Khateeb,2021)]. These techniques
are particularly useful for solving problems, where the
solution is not well-defined or where there may be a
large number of variables to consider. FS methods
aim to exclude features that are unrelated and
redundant while retaining the salient ones. This
process not only enhances overall performance but
also reduces data dimensionality, resulting in a lower
cost of classification by decreasing the training time
required to build less complex ML models (Al-
Shourbaji,2023). On the other hand, using all features
in the model increases computational overhead,
training and testing times, storage requirements, and
Al-Khateeb, M. O. and Douik, A.
Reptile Search Algorithm Based Feature Selection Approach for Intrusion Detection.
DOI: 10.5220/0013166800003890
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 17th International Conference on Agents and Artificial Intelligence (ICAART 2025) - Volume 3, pages 639-645
ISBN: 978-989-758-737-5; ISSN: 2184-433X
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.
639

error rate of ML model due to irrelevant features
confusing with the relevant ones.
Metaheuristic (MH) algorithm, also known as a
MH optimization algorithm, is a general FS
algorithmic framework that can be used to find
optimal solutions in a wide range of problem domains
(Fong,2016). These algorithms are designed to solve
complex optimization problems, where traditional
approaches may be insufficient. They are typically
inspired by natural processes like evolution, swarm
behaviour, and other complex systems (Xu,2014).
They use these models to develop search strategies
that can efficiently navigate complex search spaces,
avoiding local optima and finding globally optimal
solutions. One of the key advantages of MH
algorithms is that they are very flexible and can be
adapted to solve a wide range of problems. They are
also often faster and more efficient than traditional
optimization techniques, making them an attractive
option for large-scale optimization problems.
Some popular examples of MH algorithms
include, genetic algorithms, Particle Swarm
Optimization (PSO) (Kennedy,1995), Grey Wolf
Optimization (GWO) (Mirjalili,2014), Multi-Verse
Optimizer (MVO) (Mirjalili,2016), Remora
Optimization Algorithm (ROA) (Jia,2021), genetic
algorithm (Holland,1992) and many others. These
algorithms have been successfully applied to a diverse
range of fields, including engineering, finance,
operations research, and many others. Recently,
Reptile Search Algorithm (RSA) (Abualigah,2022),
shows a great potential as a FS method and it can pick
Optimal Feature Subset (OFS) effectively. This paper
aims to investigate FS method using RSA for ID. To
assess RSA's capabilities in determining OFS, five
publicly available ID datasets and various quantitative
evaluation measures are used. Four FS methods, PSO,
GWO, MVO, and ROA, are implemented to compare
RSA's efficiency in ID system.
The organization of remaining paper is as follows:
Section 2, briefly describes the RSA and datasets used.
Section 3, describes evaluations measurements to
evaluate RSA method. Section 4 discusses the
experimental analysis and results. Section 5 concludes
the paper.
2 METHODS AND MATERIALS
2.1 Reptile Search Algorithm (RSA)
RSA is a new method inspired by the hunting
behaviour of Crocodiles proposed by
(Abualigah,2022) in 2022. A set of candidate N
crocodiles 𝑥
,
each having random position in the
search space are initialized as follows:
𝑥
,
=𝑟𝑎𝑛𝑑
∈(,)
∗𝑈𝐵

−𝐿𝐵

+𝐿𝐵

𝑖
∈
{
1,…,𝑁
}
𝑎𝑛𝑑
𝑗
∈{1,…,𝑀}
(1)
where 𝐿𝐵

and 𝑈𝐵

are the lowest and highest
values of the 𝑗𝑡ℎ feature, 𝑟𝑎𝑛𝑑
∈(,)
generates a
number randomly in the range [0, 1] following a
uniform distribution, and M is feature dimensionality
in the dataset.
For crocodile food search, two distinct strategies,
exploration and exploitation, are employed. These
strategies are sequentially implemented over four
stages within the maximum iteration limit. In the
initial half of these stages, the algorithm leverages the
crocodile's encircling behaviour, incorporating both
high and belly walking movements, to facilitate search
space exploration. It can be formulated as:
𝑥
,
(
𝑔+1
)
=


−𝑛
,
(
𝑔
)
.𝛾 .𝐵𝑒𝑠𝑡

(
𝑔
)

−

𝑟𝑎𝑛𝑑
∈

,

.𝑅
,
(
𝑔
)

, 𝑔≤
𝑇
4
𝐸𝑆
(
𝑔
)
.𝐵𝑒𝑠𝑡

(
𝑔
)
.𝑥

∈

,

,
, 𝑔≤
2𝑇
4
𝑎𝑛𝑑 𝑔>
𝑇
4
(2)
where, for
𝑔𝑡ℎ iteration, the best position for 𝑔𝑡ℎ
feature is 𝐵𝑒𝑠𝑡

(𝑔), the hunting operator 𝑛
,
is
calculated as in Eq. (3), and the parameter for
Evolutionary Sense 𝐸𝑆(𝑔) is calculated as in Eq. (7).
ES parameter decreases over the iterations between 2
to −2. Finally, the exploration accuracy is controlled
by setting parameter 𝛾 as 0.1. The search region is
continuously decreased by parameter 𝑅
,
, calculated
as in Eq. (6). A crocodile is randomly selected by
𝑟𝑎𝑛𝑑
∈

,

to update position towards best position.
𝑛
,
=𝐵𝑒𝑠𝑡

(
𝑔
)
×𝑃
,
(3)
The normalized difference 𝑃
,
between 𝑖𝑡ℎ
crocodile's the 𝑖𝑡ℎ feature position and crocodile's
average position. It is computed as:
𝑃
,
=𝜃+
𝑥
,
−𝜇(𝑥

)
𝐵𝑒𝑠𝑡

(
𝑔
)
×𝑈𝐵

−𝐿𝐵

+𝜖
(4)
where the sensitive of the exploration is controlled
by the parameter 𝜃, while 𝜖 maintains the floor value.
𝜇
(
𝑥

)
=
1
𝑛
𝑥
,


(5)
𝑅
,
=
𝐵𝑒𝑠𝑡

(
𝑔
)
−𝑥
(
∈

,

,)
𝐵𝑒𝑠𝑡

(
𝑔
)
+𝜖
(6)
𝐸𝑆
(
𝑔
)
=2×𝑟𝑎𝑛𝑑
∈{,}
×1−
1
𝑇

(7)
ICAART 2025 - 17th International Conference on Agents and Artificial Intelligence
640

where the value 2 acts as a multiplier to provide
correlation values in the range of [0, 2], and
𝑟𝑎𝑛𝑑
∈
{
,
}
is a random integer between [−1, 1].
The search space is completely exploited by
implementing hunting coordination and cooperation
of crocodiles. It can be formulated as:
𝑥
,
(
𝑔+1
)
=

𝑟𝑎𝑛𝑑
∈

,

.𝐵𝑒𝑠𝑡

(
𝑔
)
.𝑃
,
(
𝑔
)
, 𝑔≤
3𝑇
4
𝑎𝑛𝑑 𝑔>
2𝑇
4
𝜖 . 𝐵𝑒𝑠𝑡

(
𝑔
)
.𝑛
,
(
𝑔
)
−𝑟𝑎𝑛𝑑
∈

,

.𝑅
,
(
𝑔
)
, 𝑔≤𝑇 𝑎𝑛𝑑 𝑔>
3𝑇
4
(8)
2.2 Datasets
Five openly available datasets commonly utilized for
intrusion detection (ID) assessment are chosen to
evaluate the efficiency of MH algorithms. These
datasets, widely acknowledged in the ID community
[(Z. Elgamal,2022), (S. Ekinci and D. Izci,2022)],
encompass Knowledge Discovery and Data Mining
Cup 1999 (KDD-CUP99) (M. Tavallaee,2009),
Network Security Laboratory KDD (NSL-KDD) (S.
Sapre,2019), University of New South Wales -
Network-Based 15 (UNSW-NB15) (A. Shiravi,2012),
Canadian Institute for Cybersecurity - Intrusion
Detection Evaluation Dataset 2017 (CIC-IDS2017)
and CIC-IDS2018 (I. Sharafaldin,2018).
Table 1 provides a detailed overview of these
datasets, including their source papers, feature counts,
and sample sizes. These characteristics are essential
for understanding the dataset's complexity and scale,
which are critical factors in assessing intrusion
detection techniques. Due to the computational
demands of iterative FS methods such as MH, FS is
evaluated using 10% examples of each ID dataset.
Importantly, this subsampling retains the original
balance between normal activities and network
attacks, ensuring a representative assessment of MH
algorithms' performance.
Table 1: Characteristics of intrusion detection datasets.
Dataset Source No. of
features
No. of
samples
KDD-
CUP99
(M. Tavallaee,2009) 43 494,020
NSL-KDD (S. Sapre,2019) 43 125,973
UNSW-
NB15
(A. Shiravi,2012) 49 540,044
CIC-
IDS2017
(I. Sharafaldin,2018) 78 2,827,87
6
CIC-
IDS2018
(I. Sharafaldin,2018) 80 1,048,57
5
3 EVALUATION METRICS
Intrusion detection using reduced features generated
by MH algorithms can be trained by ML models.
These models can be assessed using various
evaluation measures are used to determine how well
an ID system is performing. Some commonly used
evaluation measures in the context of ID are as
follows:
True Positive (TP): It represents the number of
instances where the ID system correctly identified an
intrusion or attack.
True Negative (TN): It represents the number of
instances where the system correctly identified non-
intrusive or normal network behaviour.
False Positive (FP): It occurs when the system
incorrectly flags normal network behaviour as an
intrusion or attack.
False Negative (FN): It occurs when the system
fails to detect an actual intrusion or attack, labelling it
as normal behaviour.
Using these basic metrics, you can calculate the
following evaluation measures:
Accuracy (AC): Accuracy measures of the overall
correctness of the ID system and is calculated as
follows:
𝐴
𝐶=
𝑇𝑃+ 𝑇𝑁
𝑇𝑃+𝑇𝑁+ 𝐹𝑁+ 𝐹𝑃
(9)
ID system, achieving a balance between P and R is
crucial. A high P indicates that when the system flags
an event as an intrusion, it is highly likely to be
accurate. A high R indicates that the system is
effective at detecting most of the actual intrusions.
Depending on the specific requirements and priorities
of the ID system, different evaluation measures may
be emphasized.
4 EXPERIMENTAL RESULTS
Evaluation of RSA's ability for identifying OFS is
conducted using five intrusion detection datasets,
comparing its performance with other MH algorithms,
including PSO (Kennedy,1995), GWO
(Mirjalili,2014), MVO (Mirjalili,2016), and ROA
(Jia,2021).
4.1 Experimental Setup
For this study, we implemented all the methods in
Python and executed them on a computer with an Intel
i7 10th generation processor, 32 GB of RAM, and
running the Windows 10 system. The parameter
Reptile Search Algorithm Based Feature Selection Approach for Intrusion Detection
641

configurations for MH algorithms are outlined in
Table 2. These settings are used based on their original
research papers.
Table 2: Parameter settings for different MH algorithms.
Method Parameters
Common
settings
Population size= 32, number of runs=20, &
number of iterations=100
PSO
𝑐

= 𝑐

= 2, 𝑤

= 0.1 and 𝑤

= 0.9
GWO
𝐶 = random in [0,2], 𝛼 & A decrease linearly
in range [2, 0] & [1, -1]
MVO
𝑊𝐸𝑃

= 1,𝑊𝐸𝑃

= 0.2, 𝛼 decreases
from 2 to 0 and p = 6
ROA
𝑙𝑑=1 and 𝛽= 2
RSA
𝛾=0.9,𝜃=0.5, UB & LB are vary based
on the features in the dataset
This setup ensures a fair and consistent evaluation of
the RSA's performance in comparison to other MH
algorithms across the all datasets.
4.2 Results and Discussion
Using the real-world datasets provided in Table 1, the
ability of RSA in selecting salient features is assessed
together with that of other MH methods.
Table 3, presented the mean and standard deviation
(STD) of fitness values of the RSA and other MH
algorithms across the five datasets. It's evident that the
RSA method was achieved the smallest average fitness
in all five datasets, indicating superior optimization
performance. The smallest STD values in all datasets
indicated better stability than other MH algorithms.
These results suggested that RSA was a competitive FS
method, as it consistently produced best fitness values
across all datasets, demonstrating its effectiveness in
optimizing fitness in the context of FS.
Table 4, provided mean and STD of the number of
features selected by MH algorithms for the five
datasets. RSA selected the fewest mean OFS for four
datasets, indicating its efficiency in FS. In the case of
KDD-CUP99, both ROA and RSA have same number
of features in OFS. RSA exhibited the lowest STD of
the number of OFS for three datasets, indicating
greater stability of FS. In UNSW-NB15 dataset, MVO
and RSA showed same STD while ROA and RSA
shared the same STD for CIC-IDS2018 dataset.
Finally, PSO had the lowest STD for CIC-IDS2017
dataset. This analysis underscored RSA's
effectiveness in selecting an OFS with lower
variability, making it a strong contender in FS tasks
across all datasets.
Table 3: Fitness for all datasets of all MH algorithms.
Datas
et
Meas
ure
Method
PSO GW
O
MV
O
RO
A
RSA
KDD-
CUP9
9
Mean 0.03
35
0.02
20
0.01
99
0.01
54
0.00
94
STD 0.00
96
0.00
93
0.00
73
0.00
78
0.00
66
NSL-
KDD
Mean 0.06
02
0.07
46
0.06
87
0.06
12
0.05
93
STD 0.00
81
0.01
02
0.00
92
0.00
93
0.00
88
UNS
W-
NB15
Mean 0.03
72
0.03
18
0.03
54
0.03
20
0.03
08
STD 0.00
75
0.00
57
0.00
52
0.00
71
0.00
49
CIC-
ID
S201
7
Mean 0.01
36
0.02
61
0.02
50
0.01
87
0.01
31
STD 0.00
60
0.00
84
0.00
66
0.00
90
0.00
82
CIC-
ID
S201
8
Mean 0.03
40
0.03
00
0.04
02
0.03
23
0.03
03
STD 0.00
72
0.00
94
0.00
93
0.00
91
0.00
61
Table 4: Number of OFS for all datasets of all MH
algorithms.
Dataset Measure
Method
PS
O
GW
O
MV
O
RO
A
RS
A
KDD-
CUP99
Mean
40 35 41 22 22
STD
5 9 6 7 3
NSL-
KDD
Mean 38 34 39 37 31
STD 4 6 5 5 3
UNSW-
NB15
Mean
33 29 37 23 25
STD
10 9 4 6 4
CIC-
IDS2017
Mean
23 63 49 25 21
STD
3 6 7 7 5
CIC-
IDS2018
Mean
45 49 71 55 43
STD
10 10 9 8 8
Table 5, compared mean and STD of accuracy of
MH algorithms for the five datasets. The proposed
RSA was outperformed the other MH methods,
consistently achieving the largest average accuracy
across four datasets. In terms of stability, the STD was
lowest for the RSA for three datasets. This indicated
that the ID systems with RSA as the FS method is
highly stable and produces reliable results. In KDD-
CUP99 dataset, GWO was achieved the lowest
accuracy STD, followed by RSA. In summary, It
highlighted RSA's effectiveness in achieving both
high mean accuracy and stability across different
datasets.
ICAART 2025 - 17th International Conference on Agents and Artificial Intelligence
642

Table 5: Accuracy for all datasets of all MH algorithms.
Dataset Measure
Method
PSO GWO MVO ROA RSA
KDD-
CUP99
Mean
0.9756 0.986 0.9895 0.9957 0.9970
STD
0.0062 0.0061 0.0032 0.0039 0.0036
NSL-
KDD
Mean 0.9481 0.9534 0.9398 0.9488 0.9528
STD
0.0059 0.1068 0.0065 0.0068 0.0052
UNSW-
NB15
Mean
0.9702 0.9747 0.9729 0.9743 0.9753
STD
0.0051 0.0051 0.0056 0.0052 0.0134
CIC-
IDS2017
Mean
0.9917 0.9884 0.9863 0.9906 0.9933
STD
0.0058 0.0052 0.0058 0.0055 0.0050
CIC-
IDS2018
Mean
0.9762 0.9812 0.9761 0.9823 0.9842
STD
0.0173 0.0190 0.0184 0.0282 0.0072
In the comparative analysis of convergence
depicted in Figure 1, after conducting 20 independent
runs for each method as recommended by (Duraibi,
2023), it becomes evident that the RSA method
consistently outperforms the other MH algorithms
across all five datasets. The RSA method was
demonstrated superior convergence rates towards
optimal solutions due to its capabilities to effectively
explore search space and visiting new regions in the
search area , underscoring its remarkable stability and
effectiveness as a FS technique for ID.
In Figure 2, we have a boxplot that displayed the
performance of multiple MH algorithms across five
different datasets. It visualized the distribution of
accuracy across the lower, middle, and upper quartile
ranges. This figure illustrated that the median
accuracy achieved by the RSA algorithm surpasses
that of the other MH algorithms for four datasets. In
case of NSL-KDD dataset, median accuracy of GWO
was slightly higher than RSA. Additionally, when
considering the upper accuracy quartile, RSA
outperformed the other algorithms in four out of the
five datasets.
(
a
)
KDD-CUP99
(
b
)
NSL-KDD
(
c
)
UNSW-NB15
(
d
)
CIC-IDS2017
(
e
)
CIC-IDS2018
Figure 1: Convergence behaviour of all MH algorithms for
ID datasets.
Reptile Search Algorithm Based Feature Selection Approach for Intrusion Detection
643

(a) KDD-CUP99
(b) NSL-KDD
(c) UNSW-
N
B15
(d) CIC-IDS2017
(e) CIC-IDS2018
Figure 2: Boxplots of accuracy of all MH algorithms for ID
datasets.
Finding the right number of features needed for the
ML task is one of the main goals of an effective FS
approach. This helps to avoid selecting either too
many or too few features. In a FS process, for instance,
picking too many features raises the possibility of
including unnecessary or redundant features, which
may result in a decline in prediction accuracy.
However, the RSA considered that the number of OFS
in their fitness function showed better performance,
and the number of the selected features were fewer
Exploration of the search space and exploitation of the
best solutions found are two conflicting objectives that
must be taken into account when using MH
algorithms. From the results provided above, RSA
demonstrated a better performance in balancing
exploration and exploitation factors with better
convergence speed as well.
5 CONCLUSION AND FUTURE
WORKS
This study presented a robust and efficient FS method
using RSA for enhancing the performance of security
solutions, particularly in the domain of ID. RSA
offered a gradient-free approach for ID and the
capabilities to avoid in getting trapped in local
optima. RSA's efficacy was thoroughly examined and
validated using five freely available ID datasets in the
ID domain. Its performance was also rigorously
compared against four other MH algorithms,
including PSO, GWO, MVO, and ROA. The results
demonstrated RSA's superiority, as it outperformed
the other MH methods across various evaluation
metrics, showcasing its capability to optimize feature
subsets effectively. In future, we plan to apply RSA
in other domains, including network attack detection
and IoT security. Also for future consideration, RSA
can be combined with another MH methods to boost
its capability and produce a novel hybrid approach for
solving complex identifications and classifications in
ID. These developments may pave the way for the
RSA to become a cornerstone method in security and
optimization research.
REFERENCES
Yadav, A., Noori, M. T., Biswas, A., & Min, B. (2023). A
concise review on the recent developments in the
internet of things (IoT)-based smart aquaculture
practices. Reviews in Fisheries Science &
Aquaculture, 31(1), 103-118.
ICAART 2025 - 17th International Conference on Agents and Artificial Intelligence
644

Khan, S. K., Shiwakoti, N., Stasinopoulos, P., & Chen, Y.
(2020). Cyber-attacks in the next-generation cars,
mitigation techniques, anticipated readiness and future
directions. Accident Analysis & Prevention, 148,
105837.
Alsoufi, M. A., Razak, S., Siraj, M. M., Nafea, I., Ghaleb,
F. A., Saeed, F., & Nasser, M. (2021). Anomaly-based
intrusion detection systems in iot using deep learning:
A systematic literature review. Applied sciences,
11(18), 8383.
Khraisat, A., Gondal, I., Vamplew, P., & Kamruzzaman, J.
(2019). Survey of intrusion detection systems:
techniques, datasets and challenges.
Cybersecurity, 2(1), 1-22.
Liu, H., & Lang, B. (2019). Machine learning and deep
learning methods for intrusion detection systems: A
survey. applied sciences, 9(20), 4396.
Al-Khateeb, M. O., Hassan, M. A., Al-Shourbaji, I., &
Aliero, M. S. (2021). Intelligent Data Analysis
approaches for Knowledge Discovery: Survey and
challenges. Ilkogretim Online, 20(5).
Al-Shourbaji, I., Kachare, P., Fadlelseed, S., Jabbari, A.,
Hussien, A. G., Al-Saqqar, F., ... & Alameen, A. (2023).
Artificial Ecosystem-Based Optimization with Dwarf
Mongoose Optimization for Feature Selection and
Global Optimization Problems. International Journal
of Computational Intelligence Systems, 16(1), 1-24.
Fong, S., Wang, X., Xu, Q., Wong, R., Fiaidhi, J., &
Mohammed, S. (2016). Recent advances in
metaheuristic algorithms: Does the Makara dragon
exist?. The Journal of Supercomputing, 72, 3764-3786.
Xu, J., & Zhang, J. (2014, July). Exploration-exploitation
tradeoffs in metaheuristics: Survey and analysis.
In Proceedings of the 33rd Chinese control
conference (pp. 8633-8638). IEEE.
Kennedy, J., & Eberhart, R. (1995, November). Particle
swarm optimization. In Proceedings of ICNN'95-
international conference on neural networks (Vol. 4,
pp. 1942-1948). IEEE.
Mirjalili, S., Mirjalili, S. M., & Lewis, A. (2014). Grey wolf
optimizer. Advances in engineering software, 69, 46-
61.
Mirjalili, S., Mirjalili, S. M., & Hatamlou, A. (2016). Multi-
verse optimizer: a nature-inspired algorithm for global
optimization. Neural Computing and Applications, 27,
495-513.
Jia, H., Peng, X., & Lang, C. (2021). Remora optimization
algorithm. Expert Systems with Applications, 185,
115665.
Holland, J. H. (1992). Genetic algorithms. Scientific
american, 267(1), 66-73.
Abualigah, L., Abd Elaziz, M., Sumari, P., Geem, Z. W., &
Gandomi, A. H. (2022). Reptile Search Algorithm
(RSA): A nature-inspired meta-heuristic
optimizer. Expert Systems with Applications, 191,
116158.
Z. Elgamal, A. Q. M. Sabri, M. Tubishat, D. Tbaishat, S. N.
Makhadmeh et al., Improved reptile search
optimization algorithm using Chaotic map and
Simulated annealing for feature selection in medical
field
, IEEE Access, vol. 10, pp. 51428–51446, 2022.
S. Ekinci and D. Izci, Enhanced reptile search algorithm
with Lévy flight for vehicle cruise control system
design, Evolutionary Intelligence, vol. 2022, pp. 1–13,
2022.
M. Tavallaee, E. Bagheri, W. Lu and A. A. Ghorbani, A
detailed analysis of the KDD CUP 99 data set, in
proceedings of IEEE conference on symposium on
Computational. Intelligence for Security and. Defense,
Ottawa, ON, Canada, pp. 1–6, 2009.
S. Sapre, P. Ahmadi and K. Islam, A robust comparison of
the KDDCup99 and NSL-KDD IoT network intrusion
detection datasets through various machine learning
algorithms, arXiv preprint, pp. 1-8, 2019.
A. Shiravi, H. Shiravi, M. Tavallaee and A. A. Ghorbani,
Toward developing a systematic approach to generate
benchmark datasets for intrusion detection, Computer.
Security, vol. 31, no. 3, pp. 357–374, 2012.
I. Sharafaldin, A. H. Lashkari and A. A. Ghorbani, Toward
generating a new intrusion detection dataset and
intrusion traffic characterization, in Proceedings of the
4th International. Conference on Information Systems.
Security and Privacy, Funchal, Portugal, vol. 1, no. 2,
pp. 108–116, 2018.
Reptile Search Algorithm Based Feature Selection Approach for Intrusion Detection
645