Customized Convolutional Neural Networks with K-Nearest Neighbor Classification System for Malware Categorization

Abstract

The rapid growth of malware systems has identified significant threats to the security of computer technology. Therefore it stimulates anti-malware providers and researchers to create new strategies that can protect users from global threats. The voluminous releases of malware that appear on the Internet identified major security threats. To address this issue, we introduce a hybrid Customized Convolutional Neural Network (CCNN) with a K-nearest neighbor (KNN) classification system that can classify variations of malware into their respective families automatically. The hybrid CCNN-KNN classification system efficiently overcomes the limitations of the existing CNN system. Thus, the proposed novel hybrid system which works on uneven nonlinear data. The experiment’s result demonstrates that CCNN-KNN is more than 99.35% precise, contrary to other CNN malware classification models. The proposed hybrid model, therefore, represents the most extensive malware classification schemes in conceptual models.