AHSWN Home · Issue Contents · Forthcoming Papers

Trust-based Anomalies Detection in Internet of Things: Optimal Neural Network for Attack Classification
Vishal Gotarane and Preeti Gupta

Recently, Internet of Things (IoT) models are attained a rising demand for automated network systems and are getting more complicated. These devices make use of wireless media for data broadcasting and thus, it is simple to attempting from attacks. Machine Learning (ML) oriented solutions are more capable to protect and identify the attacks in anomalous states. This research aims to employ a novel trust-oriented anomaly detection scheme in the IoT. This depends on building the proposed trust management by maintaining the trust score on “drop attack, replay attack, tamper attack, as well as Multiple-Max Attack (MMA)”. Based on this trust evaluation, the maliciousness of the nodes in the IoT is classified by the ML algorithm. For this attack classification, this study uses an optimized Neural Network (NN) approach. For making an accurate classification, the weights of NN are optimized using a new approach termed as Levy Based Moth Flame Algorithm (LB-MFO). Finally, the superiority of the suggested technique over existing models is investigated with respect to varied metrics.

Keywords: Attack detection; Trust score; Replay attack; Optimized NN; LB-MFO algorithm