Autoencoder based Architecture for Mitigating phishing URL attack in the Internet of Things (IoT) using Deep Neural Networks

One of the serious emerging threats on the
internet is a Phishing URL attack, wherein the attackers try to
steal the login credentials like username, passwords, or even
bank details of the user by creating fake websites then, get into
the actual website then try to corrupt or delete the information.
According to a survey conducted in 2016 at least two-thirds of
people were exposed to these phishing attacks so, these types of
attacks must be taken seriously to avoid losses occurred by these
cyber threats. There are some existing defense mechanisms like
a firewall, but it is not effective in preventing different forms of
attacks, since it works on a set of predefined rules, it identifies
and blocks only those predefined attacks. Defending mechanism
that itself able to learn the possibility of attacks so that, it not
only identifies and blocks the existing forms of attacks but also
any other forms of future attacks must be needed. Our proposed
method uses Deep Neural Networks (DNN) model combined
with autoencoders, to detect and block phishing websites in the
network layer. The model is used to feature reduction and
trained with the help of a dataset received from autoencoder,
which consists of a set of legitimate websites and phishing
websites. This architecture shows the accuracy of about 98.18%
for Multiclassification Dataset. Since the memory, availability in
sensor nodes is less only limited processing of data can be done
security algorithms are implemented in network layers.