Malware Detection Using a Novel Machine Learning Dynamic Ensemble Classification Approach

As the landscape of malware threats evolves., the need for effective detection methods becomes paramount. This study evaluates the performance of five ML algorithms Random Forest., SVM., DT., KNN., and Logistic Regression. A malware dataset from Kaggle was collected. The dataset., featuring a diverse array of features extracted from windows malware samples undergoes preprocessing and division into training and testing sets. Each algorithm is individually scrutinized for its malware detection capabilities., with metrics accuracy and recall. Later conventional ensemble learning techniques stacking classifier and voting classifier applied and achieved 98% and 97% accuracy. Subsequent to this initial analysis., a novel Dynamic Ensemble Classifier is introduced., dynamically selecting classifiers based on varying data characteristics. The proposed work compares the dynamic ensemble's performance against individual classifiers., demonstrating superior metrics. The accuracy achieved with proposed method is 99%. The dynamic ensemble exhibits adaptability to changing input data characteristics., offering a flexible and robust approach to malware detection. Experimental results showcased the efficacy of the dynamic ensemble., providing valuable insights for enhancing malware detection accuracy.