landscape, necessitating innovative approaches for its detection and classification. The study underscores the significance of feature fusion, amalgamating diverse attributes from various sources to encapsulate both static and dynamic facets of malware. Traditional single-feature methods exhibit limitations in precision, motivating the exploration of multiple characteristics for fusion and employing a unified learning algorithm for classifying malware families. The research methodology involves meticulous feature extraction, followed by the utilization of KNN, XGBoost, DecisionTree, and random forest algorithms for classification, utilizing the most critical features. Experimental results underscore the significant improvement in classification accuracy compared to conventional methods, effectively reducing false positives fusion improves malware classification accuracy by 99.11% using dynamic features, 97.31% using static features, and 99.88% using a hybrid analysis compared to the conventional method. Moreover, the study focuses on merging Convolutional Neural Network (CNN) deep learning models with feature fusion specifically for Portable Executables (PE) files, achieving a remarkable accuracy of 99.18% in discerning between benign and malicious software. This synthesis of deep learning and feature fusion remarkably fortifies mal-ware classification efficacy, offering a potent solution to combat evolving cyber threats