Security Challenges for Topology Maintenance of MANET by Trust Evaluation Mechanism

Protecting the security, privacy, and accessibility of an NIDS, or Network Intrusion Detection System, is a must. Among the various commercial applications of machine learning methods, deep learning stands out as the most significant. An efficient machine learning-based intrusion detection system is necessary for dealing with hostile activity in networks. In this study, we offer a PCA- and Residual Network-based intrusion detection system that is both quick and reliable. Dimensionality reduction is achieved by using PCA. We next create a ResNet and investigate how changing the number of primary components
$(k)$
and the time-related parameter size (time-step t) affects ResNet's classification accuracy (CA) on the UNSW_NBI5 dataset. We present a quick optimization approach for choosing
$k$
and
$t$
. Based on experimental results, our strategy achieves over
$98{{\%}}$
accuracy in binary-classification with a rate of false alarms of less than
$1.8{{\%}}$
and a duration of fewer than
$400{\mathrm{s}}$
in the training time. The accuracy for many classes is over
$86{{\%}}$
.