Botnet detection in internet of things using stacked ensemble learning model

Area of Science:

• Cyber Security
• Machine Learning
• Network Intrusion Detection

Background:

• Botnets pose a significant cyber security threat, enabling activities like cyber-attacks, spamming, and data theft.
• Existing botnet detection methods are insufficient, necessitating advanced solutions.
• The UNSW-NB15 dataset is crucial for evaluating cyber-attack detection on IoT networks.

Purpose of the Study:

• To propose a novel stacking classifier, KSDRM, for enhanced botnet detection.
• To improve the accuracy and predictive performance of botnet detection systems.
• To evaluate the effectiveness of machine learning techniques in identifying botnet attacks.

Main Methods:

• A stacking classifier (KSDRM) was developed, integrating K-nearest neighbor, support vector machine, decision tree, random forest, and multilayer perceptron.
• Logistic regression was employed as a meta-learner to combine base classifier predictions.
• Label encoding was used to transform categorical features into numerical data for machine learning models.

Main Results:

• The KSDRM model achieved 99.99% accuracy during training and 97.94% during testing.
• K-fold cross-validation demonstrated high average accuracy, ranging from 99.87% to 99.89%.
• The model effectively captured complex patterns characteristic of botnet attacks.

Conclusions:

• The KSDRM model is a highly effective method for identifying botnet-based cyber attacks.
• The proposed approach significantly enhances cyber security controls.
• The findings contribute to strengthening network defenses against dynamic cyber threats.