•  
  •  
 

Document Type

Article

Keywords

Intrusion Detection System, Cybersecurity, Deep Learning, Machine Learning, Internet of Things

Abstract

The Internet of Things (IoT), as a network of connected devices, enhances modern life but also introduces significant security vulnerabilities. Addressing these challenges requires intelligent and adaptive cybersecurity systems to ensure secure communication and protection against emerging threats. Among these systems, intrusion detection systems (IDSs) play a vital role in safeguarding IoT environments by continuously monitoring network traffic, detecting abnormal activities, and identifying or preventing unauthorized access and denial-of-service (DoS) attacks. However, despite their importance, IDSs face several limitations, including high false positive and false negative rates, delayed response times to security incidents, and substantial consumption of device resources. This paper proposes a framework for designing and implementing a hybrid for distributed intrusion detection system (DIDS) that combines traditional machine learning and advanced deep learning techniques. The model uses the random forest (RF) algorithm for feature selection (FS) and principal component analysis (PCA) for dimensionality reduction. Additionally, it integrates enhanced deep learning (DL) approaches, including an improved density peak clustering (DPC) algorithm for optimized feature representation and an enhanced long short-term memory (LSTM) algorithm for classification and model training. The proposed model is evaluated on the CICIOT2023 dataset, which reflects realistic network communication behavior alongside synthetically generated attack activities. The experimental results demonstrate a significant improvement in detection accuracy, achieving a detection rate of 97.88% while maintaining efficient resource consumption—making the system suitable for distributed deployment to monitor network traffic and generate alerts in the event of an attack.

Download

Share

COinS