This paper aims to address audio spoofing and deepfake attacks within the ASVspoof 5 Track 1 open condition. We have adopted the WavLM-AASIST, WavLM-LCNN9, and WavLMResNet18-SA models for spoofing detection. Among these, the WavLM-ResNet18-SA model demonstrated superior performance. Consequently, utilizing WavLM-ResNet18-SA, we conducted extensive ablation studies to optimize the system’s performance. These studies focused on adjusting the loss functions, normalization, data augmentation, and training dataset strategies to further enhance the model's performance. To enhance detection capabilities, we implemented a fusion strategy using variations of WavLM-ResNet18-SA, increasing the training datasets. This fusion strategy significantly improved detection accuracy, achieving a minDCF of 0.01793 and an EER of 0.648% in development set. In the final evaluation, this approach demonstrated its efficacy with a minDCF of 0.2021 and an EER of 7.01%. The results highlight the performances of our targeted enhancements and fusion techniques in handling the challenges posed by evolving spoofing and deepfake scenarios.