Improved LSTM-Based Anomaly Detection Model with Cybertwin Deep Learning to Detect Cutting-Edge Cybersecurity Attacks

Anomalies in the time series may indicate future faults—real-time system state monitoring and early alerting demand novel computational anomaly detection methods. Internet of Things (IoT) devices depend significantly on intrusion detection systems (IDS) for cybersecurity (CyberSec). Today's network security platform requires deep learning methods for dealing with complex data and distinct IDS, but current methods are insufficient. The key feature of this proposed work includes a deep learning-based Cybertwin-improved long short-term memory-anomaly detection (DL-Cyberwin-Improved LSTM-AD) model for business solutions that may achieve more prediction accuracy for IoT devices. This model analyses attacks against the Cybertwin-neural network to determine a novel model's absolute error rate threshold. In order to measure the performance of the classifiers, the CSE-CIC-IDS-2018 dataset was investigated. This paper integrates the processed data within this proposed model using the time series analysis capability of this model. A high true positive rate (TPR) of 98.19% and a low false positive rate (FPR) of 0.56% obtained using this model demonstrate the practicality of the proposed model. The test dataset assesses the model based on key metrics, including accuracy, precision, F1-score, TPR, FPR, and ROC-AUC.