International journal of Computer Networks & Communications,
Год журнала:
2024,
Номер
16(4), С. 61 - 85
Опубликована: Июль 29, 2024
The
present
research
investigates
how
to
improve
Network
Intrusion
Detection
Systems
(NIDS)
by
combining
Machine
Learning
(ML)
and
Deep
(DL)
techniques,
addressing
the
growing
challenge
of
cybersecurity
threats.
A
thorough
process
for
data
preparation,
comprising
activities
like
cleaning,
normalization,
segmentation
into
training
testing
sets,
lays
framework
model
evaluation.
study
uses
CSE-CIC-IDS
2018
LITNET-2020
datasets
compare
ML
methods
(Decision
Trees,
Random
Forest,
XGBoost)
DL
models
(CNNs,
RNNs,
DNNs,
MLP)
against
key
performance
metrics
(Accuracy,
Precision,
Recall,
F1-Score).
Decision
Tree
performed
better
across
all
measures
after
being
fine-tuned
with
Enhanced
Particle
Swarm
Optimization
(EPSO),
demonstrating
model's
ability
detect
network
breaches
effectively.
findings
highlight
EPSO's
importance
in
improving
classifiers
cybersecurity,
proposing
a
strong
NIDS
high
precision
dependability.
This
extensive
analysis
not
only
contributes
arena
providing
road
robust
intrusion
detection
solutions,
but
it
also
proposes
future
approaches
combat
changing
landscape
Journal Of Big Data,
Год журнала:
2024,
Номер
11(1)
Опубликована: Авг. 4, 2024
Abstract
As
the
number
and
cleverness
of
cyber-attacks
keep
increasing
rapidly,
it's
more
important
than
ever
to
have
good
ways
detect
prevent
them.
Recognizing
cyber
threats
quickly
accurately
is
crucial
because
they
can
cause
severe
damage
individuals
businesses.
This
paper
takes
a
close
look
at
how
we
use
artificial
intelligence
(AI),
including
machine
learning
(ML)
deep
(DL),
alongside
metaheuristic
algorithms
better.
We've
thoroughly
examined
over
sixty
recent
studies
measure
effective
these
AI
tools
are
identifying
fighting
wide
range
threats.
Our
research
includes
diverse
array
cyberattacks
such
as
malware
attacks,
network
intrusions,
spam,
others,
showing
that
ML
DL
methods,
together
with
algorithms,
significantly
improve
well
find
respond
We
compare
methods
out
what
they're
where
could
improve,
especially
face
new
changing
cyber-attacks.
presents
straightforward
framework
for
assessing
Methods
in
threat
detection.
Given
complexity
threats,
enhancing
regularly
ensuring
strong
protection
critical.
evaluate
effectiveness
limitations
current
proposed
models,
addition
algorithms.
vital
guiding
future
enhancements.
We're
pushing
smart
flexible
solutions
adapt
challenges.
The
findings
from
our
suggest
protecting
against
will
rely
on
continuously
updating
stay
ahead
hackers'
latest
tricks.
IEEE Access,
Год журнала:
2024,
Номер
12, С. 6100 - 6116
Опубликована: Янв. 1, 2024
Supervisory
Control
and
Data
Acquisition
(SCADA)
systems
are
crucial
for
modern
industrial
processes
securing
them
against
increasing
cyber
threats
is
a
significant
challenge.
This
study
presents
an
advanced
method
bolstering
SCADA
security
by
employing
modified
hybrid
deep
learning
model.
A
key
innovation
in
this
work
integrating
the
Self-similarity
Hurst
parameter
into
dataset
alongside
CNN-LSTM
model,
significantly
boosting
Intrusion
Detection
System's
(IDS)
capabilities.
The
parameter,
which
quantifies
self-similarity
dataset,
instrumental
detecting
anomalies.
Our
in-depth
analysis
of
CICIDS2017
sheds
light
on
contemporary
attack
patterns
network
traffic
behaviors.
architecture
was
substantially
altered
adding
multiple
convolutional
layers
with
progressively
filters,
batch
normalization
stable
training,
dropout
regularization.
Principal
Component
Analysis
(PCA)
applied
dimensionality
reduction,
thereby
optimizing
dataset.
Test
results
demonstrate
superior
performance
model
incorporating
achieving
95.21%
accuracy
82.59%
recall,
surpassing
standard
inclusion
marks
substantial
advancement
identifying
emerging
threats,
while
architectural
improvements
to
led
more
robust
accurate
intrusion
detection
control
settings.
Journal of Cloud Computing Advances Systems and Applications,
Год журнала:
2024,
Номер
13(1)
Опубликована: Окт. 21, 2024
Abstract
Cybersecurity
threats
have
become
more
worldly,
demanding
advanced
detection
mechanisms
with
the
exponential
growth
in
digital
data
and
network
services.
Intrusion
Detection
Systems
(IDSs)
are
crucial
identifying
illegitimate
access
or
anomalous
behaviour
within
computer
systems,
consequently
opposing
sensitive
information.
Traditional
IDS
approaches
often
struggle
high
false
positive
rates
ability
to
adapt
embryonic
attack
patterns.
This
work
asserts
a
novel
Hybrid
Adaptive
Ensemble
for
(HAEnID),
an
innovative
powerful
method
enhance
intrusion
detection,
different
from
conventional
techniques.
HAEnID
is
composed
of
string
multi-layered
ensemble,
which
consists
Stacking
(SEM),
Bayesian
Model
Averaging
(BMA),
Conditional
(CEM).
combines
best
these
three
ensemble
techniques
ultimate
success
considerable
cut
alarms.
A
key
feature
adaptive
mechanism
that
allows
components
change
over
time
as
traffic
patterns
vary
new
appear.
way,
would
provide
adequate
protection
vectors
change.
Furthermore,
model
interpretable
explainable
using
Shapley
Additive
Explanations
(SHAP)
Local
Interpretable
Model-agnostic
(LIME).
The
proposed
on
CIC-IDS
2017
achieves
excellent
accuracy
(97-98%),
demonstrating
effectiveness
consistency
across
various
configurations.
Feature
selection
further
enhances
performance,
BMA-M
(20)
reaching
98.79%
accuracy.
These
results
highlight
potential
accurate
reliable
and,
hence,
state-of-the-art
choice
explainability.
