Infrared small target detection is an important computer vision task that involves the identification and location of tiny targets in infrared images. However, due to the lack of effective information such as shape and texture in small targets in infrared images, and the generally complex background, small target detection in infrared images is full of challenges. To address the above challenges, this paper proposes a deep learning method, DifRankNet, which significantly improves the performance of infrared small target detection through multiple practical modules. Specifically, it includes a multi-path collaborative dilated differential convolution (DDC) module, a nonlinearly enhanced region rank-aware attention (ReRank) module, and a dynamically adaptive differential selection feature fusion (DSFF) module. The DDC module adopts a multi-branch feature extraction strategy to capture feature information of different scales and levels. The ReRank module adopts a nonlinear Top-K selection process for the adaptive partition block to retain the most prominent responses, prevent the dilution of the target signal, and maintain a constant complexity. The DSFF module provides adaptive feature integration, enhancing the model’s ability to distinguish real targets from false alarms. Extensive experimental results on multiple public datasets show that the proposed DifRankNet has excellent performance and outperforms other traditional models and deep learning models.