Abstract
Malware continues to pose a serious threat to modern computing systems, requiring effective and reliable detection mechanisms. Traditional malware detection approaches often rely on a single feature representation, which may fail to capture the diverse characteristics of malicious software. To address this limitation, this paper proposes a multi-view deep learning framework for malware detection using opcode-based static analysis. The proposed approach exploits two complementary representations of malware samples: opcode sequences modeled using recurrent neural networks to capture execution behavior, and opcode statistical features learned through fully connected neural networks to represent global structural characteristics. Multiple machine learning and deep learning models are first evaluated using individual views to establish baseline performance. A multi-view deep learning model is then introduced by jointly learning and fusing representations from both views. Experimental results on a publicly available malware opcode dataset demonstrate that the proposed multiview framework consistently outperforms single-view models, achieving improved classification performance. The findings confirm that integrating complementary opcode representations enhances the robustness and effectiveness of malware detection systems.