In
the
digital
era,
SQL
injection
(SQLi)
attacks
on
web
applications
pose
significant
threats
to
data
integrity
and
security.
While
traditional
methods
such
as
signature-based
anomaly-based
detections
have
some
limitations,
this
research
explores
application
of
neural
networks
in
countering
these
attacks.
Specifically,
evaluates
performance
three
primary
network
architectures:
Artificial
Neural
Networks
(ANNs),
Convolutional
(CNNs),
Recurrent
(RNNs)
for
SQLi
attack
detection.
The
methodology
involves
converting
text-based
queries
into
numeric
values
suitable
compatible
with
networks,
using
Term
Frequency-Inverse
Document
Frequency
(TF-IDF),
tokenization,
padding.
Results
show
that
CNN
outperforms
almost
all
metrics,
RNNs
following
closely
ANNs
achieving
lower
results.
IEEE Access,
Год журнала:
2023,
Номер
11, С. 135507 - 135516
Опубликована: Янв. 1, 2023
Structured
Query
Language
(SQL)
injection
attacks
represent
a
critical
threat
to
database-driven
applications
and
systems,
exploiting
vulnerabilities
in
input
fields
inject
malicious
SQL
code
into
database
queries.
This
unauthorized
access
enables
attackers
manipulate,
retrieve,
or
even
delete
sensitive
data.
The
through
underscores
the
importance
of
robust
Artificial
Intelligence
(AI)
based
security
measures
safeguard
against
attacks.
study's
primary
objective
is
automated
timely
detection
AI
without
human
intervention.
Utilizing
preprocessed
46,392
queries,
we
introduce
novel
optimized
approach,
Autoencoder
network
(AE-Net),
for
automatic
feature
engineering.
proposed
AE-Net
extracts
new
high-level
deep
features
from
textual
data,
subsequently
machine
learning
models
performance
evaluations.
Extensive
experimental
evaluation
reveals
that
extreme
gradient
boosting
classifier
outperforms
existing
studies
with
an
impressive
k-fold
accuracy
score
0.99
detection.
Each
applied
approach's
further
enhanced
hyperparameter
tuning
validated
via
cross-validation.
Additionally,
statistical
t-test
analysis
assess
variations.
Our
innovative
research
has
potential
revolutionize
attacks,
benefiting
specialists
organizations.
Security and Privacy,
Год журнала:
2024,
Номер
7(3)
Опубликована: Янв. 18, 2024
Abstract
Cloud
computing
has
revolutionized
the
way
IT
industries
work.
Most
modern‐day
companies
rely
on
cloud
services
to
accomplish
their
day‐to‐day
tasks.
From
hosting
websites
developing
platforms
and
storing
resources,
tremendous
use
in
modern
information
technology
industry.
Although
an
emerging
technique,
it
many
security
challenges.
In
structured
query
language
injection
attacks,
attacker
modifies
some
parts
of
user
still
sensitive
information.
This
type
attack
is
challenging
detect
prevent.
this
article,
we
have
reviewed
65
research
articles
that
address
issue
its
prevention
detection
Traditional
Networks,
which
11
are
related
general
rest
54
specifically
web
security.
Our
result
shows
Random
Forest
accuracy
99.8%
a
Precision
rate
99.9%,
worst‐performing
model
Multi‐Layer
Perceptron
(MLP)
SQLIA
Model.
For
recall
value,
performs
best
while
TensorFlow
Linear
Classifier
worst.
F1
score
Forest,
MLP
most
diminutive
performer.
SQL
Injection
(SQLi)
persists
as
a
major
threat
to
.NET
applications
since
attackers
can
inject
harmful
code
into
databases
for
database
manipulation
purposes.
The
presence
of
this
vulnerability
leads
hackers
gaining
access
unauthorized
data
and
causing
system
integrity
failure
while
resulting
in
lost
which
threatens
organizations
utilizing
these
applications.
Signature-based
detection
systems
demonstrate
limited
effectiveness
when
it
comes
detecting
contemporary
or
innovative
SQLi
attacks
that
create
new
patterns.
Artificial
Intelligence
through
anomaly
technology
provides
capable
defensive
solution
overcome
particular
challenge.
normal
behavior
patterns
queries
inside
become
manageable
AI
machine
learning
algorithms
detect
abnormal
signal
attack
vulnerabilities.
research
introduces
specific
AI-based
designed
application
environments.
Our
method
begins
with
collecting
query
logs
then
performing
preprocessing
before
extracting
important
features
are
used
train
model
between
valid
hostile
queries.
process
relies
on
an
RNN
autoencoder
understands
sequences
thus
identifying
anomalous
related
injection.
Experimental
testing
shows
the
proposed
reaches
high
precision
alongside
minimal
false
alarms
recognized
well
unrecognized
attacks.
security
position
becomes
more
robust
implementation
protecting
against
current
future
XML
and
SQL
injection
attacks
are
occurring
very
frequently
nowadays
as
developers
lack
major
security
measures
awareness
for
the
purpose
of
securing
web
applications
documents.
Many
factors
responsible
these
types
vulnerabilities
main
reasons
behind
all
like
urbanisation,
environmental
degradation
some
market
conditions
in
lead
to
such
attacks.
These
may
also
allow
attackers
gain
advantage
overlook
front
end
application
by
taking
input
fields
usernames,
passwords,
important
credentials.
This
research
paper
covers
detailed
study
occurrence
on
development
applications,
detection
mitigation
policies
employed
literature.
The
machine
learning
based
model
is
presented
this
case
detecting
discussed.
classification
compared
using
metrics
accuracy,
precision,
Specificity
F-1
Score.
