Intrusion Detection System Using Adaptive Learning Factor Based Sandpiper Optimization Algorithm and Bi-Directional Gated Recurrent Unit

Attacks on Internet of Things (IoT) networks are currently the most pressing issues challenging the modern society.An Intrusion Detection System (IDS) must be present to detect and reduce hostile attacks within IoT.However, the nature of network information and legitimate traffic from malicious activity lead to false positives, and unknown attacks that make the learning tasks very complex.In this research, proposed Adaptive Learning Factor based Sandpiper Optimization Algorithm (ALF-SOA) and Bi-Directional Gated Recurrent Unit (Bi-GRU) technique to detect and classify network attacks.Initially, data is obtained from two datasets namely, CICIDS2018 and CICIDS2019.The pre-processing methods like Min-Max normalization involve numerical values and adjusting mean and standard deviation to reduce sensitivity.The feature selection using ALF-SOA algorithm helps improve the efficiency and effectiveness of IDS by reducing false positives and enhancing the detection accuracy.The classification using Bi-GRU model is robust to variations in input data and improves model reliability, especially in complex data.The proposed ALF-SOA and Bi-GRU technique is evaluated using the CICIDS2018 and CICIDS2019 datasets, achieving higher accuracies of 99.90% and 99.91%, respectively.This approach is more effective than Convolutional Neural Network -Recurrent Neural Network (CNN-RNN), and Explainable Neural Network (XNN).