Malware Detection with a Hybrid Architecture Incorporating Static and Dynamic Features
Otmane Stit, Ali Yahyaouy, Jamal Riffi, Khalid Alaoui Zidani, Hamid Tairi · 2025
The rapid evolution of malware necessitates robust detection mechanisms capable of identifying both known and novel threats. Traditional malware detection techniques, such as signature-based and heuristic methods, struggle with the increasing complexity of malware obfuscation and evasion tactics. To address these challenges, this study proposes a hybrid architecture that integrates static opcode sequences and dynamic behavioral data for enhanced malware detection. The proposed model employs Long Short-Term Memory (LSTM) networks to capture sequential patterns in opcode data, Variational Autoencoders (VAEs) for feature compression and anomaly detection. By combining static features, such as opcodes and metadata, with dynamic features like API call logs and system interactions, the model achieves significant improvements in accuracy and robustness. This paper highlights the efficacy of integrating deep learning architectures with feature fusion for robust and interpretable malware detection, offering a foundation for future real-world applications.