Intrusion detection via optimal tuned LSTM model with trust and risk level evaluation

Abstract

Different heterogeneous wireless sensor networks (WSNs) linked with the cloud platform make up a sensor cloud framework. The optimal cluster head (CH) is chosen from among the SNs in this work’s introduction of the IDS model, in which the SNs with the highest energy are given priority as the CH. In particular, the energy, latency, QoS, inter-cluster distance, and intra-cluster distance are taken into account when choosing the CH. Additionally, the suggested self updated CA optimisation (SU-COA) aids in the choosing. An improved LSTM model identifies the existence of intrusions in the network. By adjusting the model’s ideal weights using the SU-COA algorithm, the detection portion is improved. For the maximum case, a less error of 1.068 is gained using LSTM+ SU-COA, while CMBO, PRO, CSO, GWO, and CA have acquired comparatively high errors of 1.097881, 1.082925, 1.090536, 1.087563 and 1.06696 for the maximum case.