Payload State Prediction Based on Real-Time IoT Network Traffic Using Hierarchical Clustering with Iterative...

Area of Science:

• Computer Science
• Cybersecurity
• Network Engineering

Background:

• Internet of Things (IoT) networks are increasingly vulnerable to sophisticated cyber threats like botnets and viruses.
• Existing network security solutions struggle with inaccurate state prediction due to communication instability and packet loss in IoT environments.
• Effective payload state prediction is crucial for early detection, warning, and rapid response to network attacks in IoT.

Purpose of the Study:

• To propose IoTGuard, a novel network payload predictor designed for real-time prediction of payload states in IoT networks.
• To address the limitations of existing state prediction schemes by developing a more accurate and efficient method.
• To enhance the security posture of IoT networks against evolving cyber threats.

Main Methods:

• Real-time extraction of application-layer payloads from IoT network traffic using a network payload separation module.
• Classification of payload states within network flows via a dedicated payload extraction module.
• Training a payload state predictor using a labeled dataset of IoT network payloads.

Main Results:

• IoTGuard demonstrated high accuracy in predicting payload states within IoT networks.
• Achieved an 86% accuracy rate in network payload prediction, outperforming the state-of-the-art method NetZob by 8%.
• Significantly reduced training time by 52.8% compared to existing methods, ensuring execution efficiency.

Conclusions:

• IoTGuard provides a more accurate and efficient solution for predicting payload states in IoT networks.
• The proposed method offers improved early warning capabilities for detecting network attacks and botnets.
• IoTGuard represents a significant advancement in securing IoT environments against prevalent cyber threats.