Machine Learning Models Sustainable Approach for Anomaly Based Intrusion Classification on Big Data

Abstract

Introduction: The exponential growth of digital data and the increasing complexity of cyber
threats demand advanced and resilient intrusion detection systems (IDS). Anomaly-based
intrusion detection, powered by machine learning (ML), has shown promise in identifying
previously unseen attacks. However, challenges remain in ensuring robustness, scalability, and
adaptability when applied to massive and dynamic datasets typical of real-world environments.
This research focuses on enhancing the robustness of machine learning models for anomaly
based intrusion classification in big data contexts.
Objectives: The researcher stated some of the following research objectives based on
sustainable approach of machine learning models: 1.To design and implement machine learning
models that effectively detects anomaly-based intrusions in large-scale network data
environments.2.To evaluate the sustainability of ML models by analyzing their computational
efficiency, energy consumption, scalability, and maintainability over time.3.To integrate big data
processing frameworks (e.g., Apache Spark, Hadoop) with ML models for efficient handling of
high-volume, high-velocity intrusion datasets.4. To enhance the robustness and adaptability of
anomaly-based intrusion detection systems (IDS) against evolving cyber threats through model
optimization and continual learning.
Methods: By integrating sustainable computing principles with machine learning algorithms
such as Random Forest, Gradient Boosting, and Deep Neural Networks, the framework
addresses the dual challenge of cyber defense and resource optimization. The model is trained
and validated using real-world big data intrusion datasets, emphasizing preprocessing
techniques, feature selection, and model robustness under diverse attack scenarios.
Results: The researcher found the results demonstrate significant improvements in detection
rates, reduced false positives, and enhanced performance metrics, highlighting the viability of
sustainable ML-driven IDS in big data environments. This research contributes to the field by
proposing a scalable, eco-conscious intrusion classification strategy aligned with modern cyber
security and sustainability goals.
Conclusions: Finally the researcher concluded that Experimental evaluations on benchmark big
data intrusion datasets demonstrate significant improvements in detection accuracy, false positive
reduction, and model stability. The results affirm the potential of strengthened ML models in
building more secure and resilient network defense systems.