Study on Feature Extraction and Classification Optimization Methods for Malware Detection Models

This paper proposes an optimization method for feature extraction and classification in malware detection. By selecting the optimal feature subset, designing a deep learning classification model, and introducing focal loss to solve the sample imbalance problem, the proposed method improves the detection accuracy. Experimental results show that the residual connection mechanism improves the model performance by 0.7%, and the feature weighting strategy further improves the detection effect by 1.2%. The optimized model achieves a detection accuracy of $\mathbf{9 8. 7 \%}$, which is 3.3% higher than the baseline model, demonstrating its good application value.