Salient Object Detection (SOD) targets to simulate the human visual attention mechanism to detect the most attention-grabbing zones in an image. However, achieving high detection performance while reducing the model’s computational cost still presents a significant challenge. To effectively alleviate this matter, we propose the Star Operation and Lightweight Multi-Scale Attention Fusion-Based Network(SLNet) for SOD. Firstly, we use StarNet as the backbone feature extraction network. Introducing the Star Operation enhances the network’s nonlinear modeling capacity, improves feature representation, and significantly reduces computational resource consumption. Secondly, we propose the Lightweight Multi-Scale Fusion Attention Module (LMFA), which employs a lightweight feature fusion strategy. This module effectively reduces computational cost while maintaining detection accuracy, simultaneously minimizing feature redundancy. Furthermore, we introduce the Feature Enhancement Module (FE), which infers attention across different dimensions to enhance feature representations without introducing additional overhead, thereby improving detection performance. Finally, we undertake rigorous analysis on a range of publicly accessible datasets. Compared to the baseline, SLNet achieves a 68% reduction in model parameters, reaching 11.064 M, and a 54.7% decrease in computational cost, with 8.548 G Flops. Moreover, SLNet outperforms existing methods in detection performance.