Online Adaboost-Based Parameterized Methods for Dynamic Distributed Network Intrusion Detection.

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Traditional network intrusion detection systems struggle with adaptability in evolving network environments.
• Distributed architectures present new challenges for effective intrusion detection.
• The need for adaptive and robust intrusion detection is paramount in modern networks.

Purpose of the Study:

• To propose novel online Adaboost-based algorithms for network intrusion detection.
• To develop a distributed intrusion detection framework for enhanced adaptability.
• To improve detection rates and reduce false alarm rates in dynamic network settings.

Main Methods:

• Developed two online Adaboost algorithms: one using decision stumps, another using Gaussian Mixture Models (GMMs).
• Proposed a distributed framework employing local Adaboost models combined into a global model using Particle Swarm Optimization (PSO) and Support Vector Machines (SVM).
• Evaluated algorithm performance against existing methods in terms of detection rate and false alarm rate.

Main Results:

• The improved online Adaboost with GMMs demonstrated a higher detection rate and lower false alarm rate compared to the traditional approach.
• Both proposed algorithms outperformed existing intrusion detection methods.
• The PSO-SVM based combination effectively created global models capable of detecting novel intrusion types without sample sharing.

Conclusions:

• The proposed online Adaboost algorithms, particularly with GMMs, offer significant improvements in network intrusion detection.
• The distributed framework effectively integrates local detection models, enhancing robustness and adaptability.
• This approach provides a scalable and efficient solution for intrusion detection in complex, distributed network architectures.