Reliable fault diagnosis for civil aircraft operating under variable working conditions is crucial for ensuring safety and operational efficiency. However, the limited availability of high-confidence fault samples poses significant challenges, including reduced inference accuracy and an increased risk of overfitting. To address these issues, we propose the Condition Diffusion-based Fault Diagnosis (CDFD) method, which employs a denoising diffusion model to generate high-confidence fault samples. Unlike conventional diffusion models that primarily focus only on sample distribution inference, CDFD integrates sample generation with decision-making optimization. This integration facilitates the generation of high-confidence data and enhances the generalization capability of the diagnostic process. Experimental results on both simulated and real-world datasets demonstrate the effectiveness of the proposed method in improving fault diagnosis performance.