| 摘要: |
| 为进一步提高物理层密钥生成过程中量化阶段的密钥一致性和可靠性，提出了一种基于距离的样本筛选策略。该策略通过测量样本与聚类中心的欧氏距离差，评估分类不确定性，剔除高不确定度样本以减少噪声干扰。仿真结果表明，在 k 均值量化算法和补偿 k 均值量化算法中引入该策略后，当单个样本量化为 5比特时，分类不一致率分别降低 8.1% 和 11.7%；量化为单比特时，分别降低 63.4% 和 89.3%。 |
| 关键词: 物理层安全 密钥生成 信道量化：k均值聚类 距离裁剪策略 |
| DOI：10.20079/j.issn.1001-893x.250114002 |
|
| 基金项目:唐山市人才项目(B202302013);唐山市科技项目(24130215C) |
|
| Improved k-Means Clustering Quantization Algorithm with Distance Trimming Strategy |
| ZHA Kun,AN Yongli,LIU Yingchao,SONG Wenfeng |
| (College of Artificial Intelligence,North China University of Science and Technology,Tangshang 063210,China) |
| Abstract: |
| A distance-based sample screening strategy is proposed to enhance key consistency during physical layer key generation.The strategy evaluates classification uncertainty by calculating the Euclidean distance difference between samples and clustering centers,then eliminates high-uncertainty samples to reduce noise interference.Simulation results show that when applied to k-Means quantization and compensated k-Means quantization,the classification inconsistency rate decreases by 8.1% and 11.7% for 5-bit quantization,and by 63.4% and 89.3% for 1-bit quantization,respectively. |
| Key words: physical layer security key generation channel quantization k-Means clustering distance trimming strategy |
《电讯技术》编辑部 