Brain tumors pose life-threatening risks by compressing and damaging brain regions, leading to significant impairments. Advanced neuroimaging techniques aid radiologists in segmenting brain tumors, but manual segmentation from neuroimages remains time-consuming and challenging due to noise, intensity and texture variations, tissue merging, and overlapping tissue intensities. Timely and accurate brain tumor segmentation is crucial, highlighting the need for automated segmentation systems. Existing methods have limitations, necessitating further investigation and improvement. In this paper, a novel method for brain tumor segmentation based on Generative Adversarial Networks (GANs) inspired by the attention mechanism is proposed. The core idea is to utilize an efficient attention-driven Residual U-Net model with ResNext50 as its backbone for the generator to produce segmented tumor masks and VGG19 as the discriminator to validate the segmentation. The accuracy of tumor segmentation is further enhanced by incorporating attention-inspired GAN. The proposed method is assessed on the Kaggle low-grade gliomas (LGG) dataset, and demonstrates commendable efficacy and superior tissue segmentation utilizing FLAIR modality when compared to state-of-the-art methods, systematically benchmarking its superiority over them.