The detection of ships using Synthetic Aperture Radar (SAR) imagery is crucial in military and civilian sectors, yet balancing speed and accuracy remains challenging. Existing lightweight models aim to reduce parameter count and computational load but often compromise accuracy, leading to increased false alarms. Complex backgrounds, especially near-shore ships against intricate backgrounds, exacerbate SAR ship detection difficulties. This paper introduces an enhanced lightweight ship detection method, GSE-ships, based on the YOLOv5s model, augmented with an attention mechanism. Initially, the proposed GhostMSENet lightweight model facilitates rapid and precise localization and detection of ships within SAR images to fulfill the requirement for real-time detection. Subsequently, integration of the improved attention mechanism enhances the overall detection efficacy of ship objects amidst complex backgrounds post-model training. Lastly, the adoption of the novel ablation function, ECIoU, promotes early convergence and augments ship object detection capabilities. Experimental findings conducted on the SAR ship detection dataset HRSID demonstrate that the enhanced GSE-ships model achieves a reduction in computational burden by 7.97% while enhancing average accuracy by 3.1% compared to YOLOv5s. These results underscore its heightened potential for real-time ship detection applications.