这是一篇计算机领域的论文摘要请改为专业、地道的英文表述。在机器学习中由于有标记数据获取成本高充分利用廉价无标记数据进行模型训练的半监督学习发展迅速且取得广泛应用。但是传统的半监督学习框架假设有标记数据和无标记数据具有相同分布事实上这一假设在许多应用场景中难以成立。当无标记数据中具有分布外的未知类数据时模型的有效性就会受到较大的影响。针对这种情况本课题设计完成了一种安全半监督学习方法：一方面采用对比

In the field of computer science, semi-supervised learning has rapidly developed and gained widespread application in machine learning due to the high cost of obtaining labeled data. However, traditional semi-supervised learning frameworks assume that labeled and unlabeled data have the same distribution, which is often not the case in many application scenarios. When the unlabeled data contains unknown class data that is outside the distribution, the effectiveness of the model is greatly affected. To address this issue, this study proposes a secure semi-supervised learning method that utilizes contrastive learning as the main framework for feature learning to fully utilize all data for better representation and out-of-distribution detection abilities. Additionally, the model's update direction is explicitly guided using gradient projection to ensure that the classification performance of known classes within the distribution is not affected. Finally, the effectiveness of the proposed method is compared with three other methods from various perspectives.