The widespread adoption of automated code generation tools has led to a rapid increase in the number of malware variants, posing significant challenges to existing defense strategies. To tackle this issue, this paper proposes a novel malware detection and classification method based on image sample reconstruction and feature enhancement. Initially, missing values in the raw data samples are addressed, and standardization is performed. The samples are then converted into images, with channel padding and fusion applied to generate reconstructed image samples. In the feature enhancement stage, a dual-branch feature extractor is employed to separately capture both image and textual features. These features are then enhanced using a channel-spatial multi-dimensional attention network, which strengthens the feature representation by focusing on the most relevant parts of the data. Experimental results demonstrate that the proposed method significantly outperforms existing approaches, effectively improving malware detection and classification performance.