Fog computing deep learning-based intrusion detection model for IOT network security

Abstract

The rapid expansion of Internet of Things (IoT) communications across fog networks has led to a significant increase in security concerns and cyber threats due to the multiplicity and diversity of violations and security vulnerabilities. Due to inadequate security, a large number of IoT devices are vulnerable to malware, security breaches, and cyberattacks. It has been challenging for conventional intrusion detection systems (IDS), which rely on signature-based detection, to identify novel and unexpected attacks. Thus, deep learning and machine learning technologies offer interesting options for IoT security. A type of deep recurrent neural network (DRNN) called bidirectional long short-term memory networks (Bi-LSTM) is proposed in this paper as an intelligent intrusion detection system (IDS) for the Internet of Things (IoT). Through the careful analysis of complex network traffic patterns, the system can detect known and novel attacks with high accuracy and low false alarm rates. The training includes neural networks, anomaly detection, rare feature extraction, and data processing, all of which are parts of the process. The model is evaluated using accuracy, recall, F1 score, and false alarm rates after being trained with 70% of industrial datasets generated by MATLAB that are similar to real data (such as CICIoT2023 and CICIDS2017) and 30% of the time for testing. The results show that while it increases intrusion detection accuracy, the proposed intrusion detection system provides high-precision security monitoring and practical software simulation. The main contributions include efficient intrusion detection, improved detection accuracy, and increased data security. Because the model works with a number of Internet of Things communication protocols, it provides a practical security solution for protecting IoT networks from advanced cyberattacks. Simulations emphasize reproducibility, and strategies are planned using a standard dataset from Internet sources. She participated in creating and evaluating a deep learning model, achieving a training accuracy of 99.991% in detecting and preventing attacks.