STID-Net: Optimizing Intrusion Detection in IoT with Gradient Descent

Area of Science:

• Cybersecurity
• Network Security
• Machine Learning

Background:

• The proliferation of Internet of Things (IoT) devices in medical and industrial sectors has amplified network vulnerabilities.
• Existing intrusion detection systems (IDS) often fail to capture complex, irregular patterns in dynamic IoT data, limiting their applicability.
• A robust and scalable IDS is crucial for securing diverse IoT applications.

Purpose of the Study:

• To propose STID-Net, a novel intrusion detection system designed to address the limitations of current methods in dynamic IoT environments.
• To enhance the detection of spatial and temporal patterns, including long-term dependencies, in network intrusion data.
• To evaluate the performance and robustness of STID-Net across different IoT application datasets.

Main Methods:

• STID-Net integrates customized convolutional kernels for spatial feature extraction and Long Short-Term Memory (LSTM) layers for temporal sequence modeling.
• An attention mechanism is incorporated to improve the detection of long-term dependencies within intrusion patterns.
• The system was experimented with Mini-Batch Gradient Descent (MBGD) and Stochastic Gradient Descent (SGD) optimizers on Internet of Medical Things (IoMT) and Industrial Internet of Things (IIoT) datasets.

Main Results:

• STID-Net achieved high accuracy, with SGD optimization yielding 98.58% on IoMT and 99.15% on IIoT datasets, surpassing MBGD optimization (97.14% and 97.85%, respectively).
• The SGD optimizer demonstrated faster convergence and better weight adjustments, proving effective for noisy datasets.
• STID-Net outperformed standalone Convolutional Neural Network (CNN) and LSTM models, showcasing its superior performance and robustness.

Conclusions:

• STID-Net demonstrates superior capability in identifying irregular patterns and long-term dependencies in dynamic intrusion data.
• The proposed model is robust and scalable for diverse IoT applications, particularly in the medical and industrial domains.
• SGD optimization enhances STID-Net's performance, making it a reliable solution for real-world network security challenges.