IoT-based anomaly detection with the hybrid activation and the attention mechanism enabled LSTM-GRU

• The research detects anomalies in IoT data using the HA2-LRU model. • HAF and HAM enable fine-grained feature extraction for accurate analysis. • HAF, HAM, LATM, and GRU together enhance anomaly detection in traffic. • Fusion of HAF and HAM integrates multiple features and captures dependencies. Network anomaly detection remains the major concern of the current world especially, in the Internet of Things (IoT) data. The data collected from the various sources are combined and the transfer of the data remains under high threat due to the intruders. Several researches have emerged to provide better outcomes but failed to include certain characteristics such as the convergence time, computational complexity, and so on. The research with Hybrid Activation and the attention mechanism enabled LSTM-GRU (HA2-LRU) is proposed to work with the challenges of the existing methods and to provide better outcomes. The model included the combination of the LSTM and the GRU that enhanced the detection accuracy through the great involvement of the features. The incorporation of the Hybrid activation function (HAF) and the Hybrid attention mechanism (HAM) into the Long-Short term memory and Grated Recurrent Unit (LSTM-GRU) added the enhanced advantages due to the highly defined feature extraction process of the research. The research efficiency is analyzed with metrics such as the accuracy, precision, recall, and F1 score that achieves 97.60%, 90.11%, 97.58%, and 93.69% with TP and 96.45%, 92.00%, 94.56%, and 93.26% with K-Fold.