Deep Neural Decision Forest (DNDF): A Novel Approach for Enhancing Intrusion Detection Systems in Network Traffic...

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Intrusion detection systems (IDSs) are critical for network security, acting as the first line of defense against cyberattacks.
• Traditional IDSs often fail to provide accurate, real-time threat detection due to evolving cyber threats and network traffic complexity.
• There is a pressing need for innovative methods to improve IDS performance in network traffic analysis.

Purpose of the Study:

• To introduce and evaluate a novel Deep Neural Decision Forest (DNDF) model for enhancing intrusion detection.
• To assess the DNDF model's effectiveness in network traffic analysis using multiple datasets.
• To demonstrate the DNDF model's superiority over existing approaches in detecting network intrusions.

Main Methods:

• Developed a Deep Neural Decision Forest (DNDF) model, integrating deep networks with classification trees for enhanced data representation learning.
• Utilized the CICIDS 2017 dataset for initial network traffic analysis and model training.
• Extended experimental evaluation to the CICIDS 2018 and a custom network traffic dataset to validate performance.

Main Results:

• The DNDF model achieved a remarkable precision of 99.96% on the CICIDS 2017 dataset.
• DNDF demonstrated superior performance compared to reference approaches in network intrusion detection.
• The model effectively created latent representations in deep layers, contributing to its high accuracy.

Conclusions:

• The Deep Neural Decision Forest (DNDF) model offers a significant advancement in intrusion detection capabilities.
• DNDF's success is attributed to improved feature representation, model optimization, and resilience to noisy/unbalanced data.
• The DNDF model presents a robust solution for enhancing network security and intrusion detection systems.