A hybrid network intrusion detection using darwinian particle swarm optimization and stacked autoencoder hoeffding...

Area of Science:

• Cybersecurity
• Data Mining
• Machine Learning

Background:

• Cyber-attack damage costs are escalating, necessitating robust methods for secure data transmission.
• Intrusion Detection Systems (IDS) are vital for identifying deviations and protecting organizational data.
• Traditional IDS methods face challenges in efficiently detecting subtle, hidden threats in large datasets.

Purpose of the Study:

• To enhance the effectiveness of Intrusion Detection Systems (IDS) through advanced data mining techniques.
• To develop a novel approach for rapid and accurate identification of sensitive information and user deviations.
• To improve the speed and accuracy of threat detection in network traffic.

Main Methods:

• Incorporation of stream data mining with an IDS framework.
• Development and application of the Stacked Autoencoder Hoeffding Tree (SAE-HT) approach.
• Utilizing Darwinian Particle Swarm Optimization (DPSO) for efficient feature selection from the NSL-KDD dataset.

Main Results:

• The proposed SAE-HT technique achieved a high accuracy of 97.7% on the NSL_KDD dataset.
• The DPSO-based feature selection effectively identified crucial network traffic characteristics.
• The SAE-HT approach demonstrated superior performance compared to existing state-of-the-art techniques.

Conclusions:

• The SAE-HT approach significantly improves IDS accuracy and detection rates.
• The integration of DPSO for feature selection enhances the efficiency of the IDS.
• This method offers a promising solution for real-time threat detection and reducing false alarm rates in cybersecurity.