Infrared small target detection is challenged by scale variation of targets and complex background noise due to the inherent characteristics of infrared imaging. While CNN-based methods have shown promising results, their limited receptive fields restrict the modeling of global context. Transformer-based solutions alleviate this limitation but incur high computational costs. To address these challenges, we propose GLKA-UNet, a global-local aware UNet with KAN attention that efficiently captures both global and local features. First, we propose a global-local feature extraction module that leverages the fast Fourier transform to extract global frequency-domain features, which are fused with spatial-domain local features for improved feature representation. Second, we design a hierarchical feature enhancement module that combines multi-scale perception and KAN Attention to improve scale adaptivity and robustness against complex background noise. Experimental results show that GLKA-UNet surpasses the SOTA model. The proposed method achieves improvements of 0.48% and 0.85% in IoU and increases of 0.78% and 1.98% in Pd on two benchmark datasets, respectively, fully validating the effectiveness of the proposed model. The code is publicly available at https://github.com/zhangxiangping677-ai/GLKA_UNet-for-IRSTD .