Automatic
Speech
Recognition
(ASR)
is
a
crucial
application
of
deep
learning
in
today's
world.
However,
ASR
systems
are
vulnerable
to
attacks
from
malicious
actors
who
can
create
adversarial
examples
fool
the
system
into
producing
incorrect
outputs.
Adversarial
example
detection
techniques
being
developed
mitigate
risk
such
attacks.
This
review
analyzes
effectiveness
various
context
systems.
It
begins
by
explaining
concept
and
their
potential
impact
on
as
well
most
used
evaluation
metrics.
Then,
it
provides
an
overview
different
types
that
be
launched
against
Upon
reading
this
review,
researchers
able
identify
existing
weakness
ASR,
evade
them
future
proposed
schemes.
In
the
complex
landscape
of
modern
networks,
necessity
Intrusion
Detection
System
(IDS)
has
become
paramount.
An
IDS
is
a
crucial
cybersecurity
tool
that
plays
pivotal
role
in
safeguarding
networks
against
wide
array
threats
and
attacks.
The
application
deep
learning
models
for
intrusion
detection
becoming
popular
among
research
communities
due
to
its
success
many
other
domains.
However,
require
significant
amount
labeled
data
achieve
effective
training.
Obtaining
can
be
challenging
costly.
To
address
it,
Deep
Transfer
Learning
(DTL)
employed.
This
introduces
an
innovative
traffic
classification
method
tailored
5G
networks.
approach
leverages
transfer
by
utilizing
pre-trained
fine-tuning
them.
We
evaluate
several
deep-learning
setting.
Inception
model
being
identified
as
top-performing
shows
improvement
approximately
10%
terms
F1-score
between
IDS-based
DTL
same
scheme
without
DTL.
International Journal of Advanced Computer Science and Applications,
Journal Year:
2024,
Volume and Issue:
15(2)
Published: Jan. 1, 2024
The
growth
of
the
internet
has
advanced
information-sharing
capabilities
and
vastly
increased
importance
global
network
security.
However,
because
new
inconspicuous
abnormal
behaviors
are
nearly
impossible
to
detect
in
massive
access
environments,
modern
intrusion
detection
systems
have
identified
a
high
rate
false-positive
(FP)
false-negative
(FN)
attacks.
To
overcome
this,
this
paper
proposes
hybrid
deep
learning
model
that
significantly
mitigates
disadvantages
consistently
imbalanced
sample
attack
data.
First,
it
resolves
data
using
random
undersampling
synthetic
minority
oversampling
techniques.
Then,
convolutional
neural
networks
(CNNs)
extract
local
spatial
features,
transformer
encoder
extracts
temporal
features.
novelty
combination
increases
recognition
accuracy
at
algorithm
level,
which
is
crucial
reducing
FPs
FNs.
was
subjected
multiclassification
testing
on
NSL-KDD
CICIDS2017
benchmark
datasets,
results
show
our
higher
classification
lower
FP
rates
than
state-of-the-art
models.
Moreover,
improves
low-frequency
Automatic
Speech
Recognition
(ASR)
is
a
crucial
application
of
deep
learning
in
today's
world.
However,
ASR
systems
are
vulnerable
to
attacks
from
malicious
actors
who
can
create
adversarial
examples
fool
the
system
into
producing
incorrect
outputs.
Adversarial
example
detection
techniques
being
developed
mitigate
risk
such
attacks.
This
review
analyzes
effectiveness
various
context
systems.
It
begins
by
explaining
concept
and
their
potential
impact
on
as
well
most
used
evaluation
metrics.
Then,
it
provides
an
overview
different
types
that
be
launched
against
Upon
reading
this
review,
researchers
able
identify
existing
weakness
ASR,
evade
them
future
proposed
schemes.
