2022 IEEE World Conference on Applied Intelligence and Computing (AIC),
Год журнала:
2023,
Номер
unknown
Опубликована: Июль 29, 2023
Efficient
malware
identification
is
essential
to
safe
the
system
resources
and
privacy
of
data
for
cybersecurity
system.
The
use
android
smartphones
has
increased
tremendously
that
attracting
various
types
attacks.
Nowadays,
writers
Artificial
Intelligence
(AI)-enabled
attack
techniques
bypass
detection
malicious
activities.
Hence,
designing
an
efficient,
effective
robust
identify
variants
remains
a
critical
problem
challenge.
However,
number
deep
learning
(DL)
models
applied
in
existing
methods
at
large
scale,
but
these
actually
lacks
interpretability
explain
contribution
each
features
Therefore,
this
paper
propose
Explainable
(XAI)
based
hybrid
Convolutional
Neural
network
(CNN)
Bi-Gated
Recurrent
Unit
(Bi-GRU)
Android
Malware
Detection
(AMD)
System
using
DL
named
as
XAI-AMD-DL.
proposed
model
evaluated
CICAndMal2019
dataset.
results
obtained
by
XAI-AMD-DL
97.98%
accuracy,
97.75
%,
97.76%,
97.75%
precision,
recall
f1score,
respectively
outperforms
models.
Sensors,
Год журнала:
2023,
Номер
23(19), С. 8153 - 8153
Опубликована: Сен. 28, 2023
The
Internet
of
Things
(IoT)
and
network-enabled
smart
devices
are
crucial
to
the
digitally
interconnected
society
present
day.
However,
increased
reliance
on
IoT
increases
their
susceptibility
malicious
activities
within
network
traffic,
posing
significant
challenges
cybersecurity.
As
a
result,
both
system
administrators
end
users
negatively
affected
by
these
malevolent
behaviours.
Intrusion-detection
systems
(IDSs)
commonly
deployed
as
cyber
attack
defence
mechanism
mitigate
such
risks.
IDS
plays
role
in
identifying
preventing
hazards
networks.
development
an
efficient
rapid
for
detection
attacks
remains
challenging
area
research.
Moreover,
datasets
contain
multiple
features,
so
implementation
feature
selection
(FS)
is
required
design
effective
timely
IDS.
FS
procedure
seeks
eliminate
irrelevant
redundant
features
from
large
datasets,
thereby
improving
intrusion-detection
system's
overall
performance.
In
this
paper,
we
propose
hybrid
wrapper-based
feature-selection
algorithm
that
based
concepts
Cellular
Automata
(CA)
engine
Tabu
Search
(TS)-based
aspiration
criteria.
We
used
Random
Forest
(RF)
ensemble
learning
classifier
evaluate
fitness
selected
features.
proposed
algorithm,
CAT-S,
was
tested
TON_IoT
dataset.
simulation
results
demonstrate
enhances
classification
accuracy
while
simultaneously
reducing
number
false
positive
rate.
Journal of Informatics Education and Research,
Год журнала:
2025,
Номер
5(1)
Опубликована: Янв. 17, 2025
A
revolutionary
chance
to
improve
consumer
engagement
exists
with
the
incorporation
of
AI.
This
proposed
delves
at
various
ways
generative
AI
may
be
used
in
digital
marketing
campaigns,
highlighting
how
it
can
change
way
customers
engage
and
content
is
made.
Personalized
content,
made
possible
by
AI's
capacity
sift
through
mountains
customer
data,
main
emphasis.
Chatbots
virtual
assistants
powered
artificial
intelligence
are
also
investigated
study
for
their
potential
offer
real-time
assistance
interactivity.
By
improving
user
experience
keeping
attention,
this
technology
encourages
more
in-depth
brand
involvement.
The
article
assesses
well
AI-powered
social
media
strategy
optimization
works.
Increased
conversion
rates,
better
engagement,
fine-tuned
tactics
results
automation.
Finally,
takes
into
account
growing
significance
optimizing
campaigns
voice
visual
searches,
where
improves
visibility
accessibility
these
new
search
methods.
In
show
that
essential
developing
tailored
each
individual,
creative,
responsive
than
ever
before.
Generative
a
game-changer
efforts
because
combines
creativity,
efficiency,
personalization
increase
client
engagement.
Sensors,
Год журнала:
2025,
Номер
25(7), С. 2288 - 2288
Опубликована: Апрель 4, 2025
As
more
IoT
devices
become
connected
to
the
Internet,
attack
surface
for
cybercrimes
expands,
leading
significant
security
concerns
these
devices.
Existing
intrusion
detection
systems
(IDSs)
designed
address
often
suffer
from
high
rates
of
false
positives
and
missed
threats
due
presence
redundant
irrelevant
information
IDSs.
Furthermore,
recent
IDSs
that
utilize
artificial
intelligence
are
presented
as
black
boxes,
offering
no
explanation
their
internal
operations.
In
this
study,
we
develop
a
solution
identified
challenges
by
presenting
deep
learning-based
model
adapts
new
attacks
selecting
only
relevant
inputs
providing
transparent
operations
easy
understanding
adoption
cybersecurity
personnel.
Specifically,
employ
hybrid
approach
using
statistical
methods
metaheuristic
algorithm
feature
selection
identify
most
features
limit
overall
set
while
building
an
LSTM-based
detection.
To
end,
two
publicly
available
datasets,
NF-BoT-IoT-v2
IoTID20,
training
testing.
The
results
demonstrate
accuracy
98.42%
89.54%
IoTID20
respectively.
performance
proposed
is
compared
with
other
machine
learning
models
existing
state-of-the-art
models,
demonstrating
superior
accuracy.
explain
model's
predictions
increase
trust
in
its
outcomes,
applied
explainable
(XAI)
tools:
Local
Interpretable
Model-agnostic
Explanations
(LIME)
Shapley
Additive
(SHAP),
valuable
insights
into
behavior.
