Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimized detection of cyber-attacks on IoT networks via hybrid deep learning models

Ahmed Bensaoud, Jugal Kalita

TL;DR

This paper addresses the challenge of detecting both known and unknown cyber-attacks in IoT networks by proposing a hybrid framework that fuses Self-Organizing Maps (SOMs), Deep Belief Networks (DBNs), and Autoencoders (AEs). The approach is further enhanced by a novel Particle Swarm Optimization (PSO) scheme that jointly tunes hyperparameters across all three models, with an objective that combines reconstruction, clustering, and hierarchical losses. Experimental results on NSL-KDD, UNSW-NB15, and CICIoT2023 show near-perfect accuracy and MCC on multiple attack classes, demonstrating strong robustness to emerging threats and dataset diversity. The work highlights the potential of hybrid unsupervised-supervised architectures, augmented with optimization and attention-based interpretability, to deliver scalable, real-time intrusion detection for dynamic IoT environments, while outlining avenues for privacy-preserving, edge-enabled, and explainable enhancements.

Abstract

The rapid expansion of Internet of Things (IoT) devices has increased the risk of cyber-attacks, making effective detection essential for securing IoT networks. This work introduces a novel approach combining Self-Organizing Maps (SOMs), Deep Belief Networks (DBNs), and Autoencoders to detect known and previously unseen attack patterns. A comprehensive evaluation using simulated and real-world traffic data is conducted, with models optimized via Particle Swarm Optimization (PSO). The system achieves an accuracy of up to 99.99% and Matthews Correlation Coefficient (MCC) values exceeding 99.50%. Experiments on NSL-KDD, UNSW-NB15, and CICIoT2023 confirm the model's strong performance across diverse attack types. These findings suggest that the proposed method enhances IoT security by identifying emerging threats and adapting to evolving attack strategies.

Optimized detection of cyber-attacks on IoT networks via hybrid deep learning models

TL;DR

This paper addresses the challenge of detecting both known and unknown cyber-attacks in IoT networks by proposing a hybrid framework that fuses Self-Organizing Maps (SOMs), Deep Belief Networks (DBNs), and Autoencoders (AEs). The approach is further enhanced by a novel Particle Swarm Optimization (PSO) scheme that jointly tunes hyperparameters across all three models, with an objective that combines reconstruction, clustering, and hierarchical losses. Experimental results on NSL-KDD, UNSW-NB15, and CICIoT2023 show near-perfect accuracy and MCC on multiple attack classes, demonstrating strong robustness to emerging threats and dataset diversity. The work highlights the potential of hybrid unsupervised-supervised architectures, augmented with optimization and attention-based interpretability, to deliver scalable, real-time intrusion detection for dynamic IoT environments, while outlining avenues for privacy-preserving, edge-enabled, and explainable enhancements.

Abstract

The rapid expansion of Internet of Things (IoT) devices has increased the risk of cyber-attacks, making effective detection essential for securing IoT networks. This work introduces a novel approach combining Self-Organizing Maps (SOMs), Deep Belief Networks (DBNs), and Autoencoders to detect known and previously unseen attack patterns. A comprehensive evaluation using simulated and real-world traffic data is conducted, with models optimized via Particle Swarm Optimization (PSO). The system achieves an accuracy of up to 99.99% and Matthews Correlation Coefficient (MCC) values exceeding 99.50%. Experiments on NSL-KDD, UNSW-NB15, and CICIoT2023 confirm the model's strong performance across diverse attack types. These findings suggest that the proposed method enhances IoT security by identifying emerging threats and adapting to evolving attack strategies.

Paper Structure

This paper contains 49 sections, 20 equations, 34 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Data Privacy and Security in IoT Networks
  4. Network Intrusion Detection in IoT
  5. Deep Learning Approaches for IoT Security
  6. Hybrid Models for Network Intrusion Detection
  7. Optimization Techniques in IoT Intrusion Detection
  8. Attention Mechanisms for Enhancing Model Interpretability
  9. Challenges in Detecting Unknown Attacks
  10. Methodology
  11. Self-Organizing Maps (SOMs)
  12. Initialization
  13. Training Process
  14. Deep Belief Networks (DBNs)
  15. Restricted Boltzmann Machines (RBMs)
  16. ...and 34 more sections

Figures (34)

  • Figure 1: Integration of SOMs, DBNs, and Autoencoders for Model Framework.
  • Figure 2: Parallel Coordinates Plot showing the distribution of Source Bytes, Rerror Rate, and Duration across different class labels.
  • Figure 3: Histogram of source bytes with KDE overlay, showing byte size distribution and insights into normal versus potentially malicious network traffic.
  • Figure 4: Pairwise relationships and distributions of selected NSL-KDD features for normal and attack traffic.
  • Figure 5: Pairwise relationships and distributions of selected UNSW-NB15 features for normal and attack traffic.
  • ...and 29 more figures