A Deep Learning-Based Hybrid Architecture for Polymorphic Malware Detection

PolyDefender introduces a hybrid deep learning framework for detecting polymorphic malware by combining convolutional neural networks (CNNs) and long short-term memory networks (LSTMs). The system performs both static and dynamic analysis on executable files, extracting byte-level patterns and behavioral API call sequences. Trained on 50,000 malware and benign samples, it achieves a 97.8% detection accuracy with minimal false positives (0.3%). It also incorporates adversarial training to defend against evasion attacks. This research offers a resilient approach for modern malware defense.