基于深度迁移学习的网络敏感信息快速辨识研究  

Research on Rapid Identification of Network Sensitive Information Based on Deep Transfer Learning

作　　者：王彩玲[1] WANG Cailing(Network Security Department of Henan Police College,Zhengzhou 450046,China)

机构地区：[1]河南警察学院网络安全系,郑州450046

出　　处：《微处理机》2025年第2期44-51,共8页Microprocessors

摘　　要：本研究旨在解决传统方法在网络敏感信息辨识中因单一特征提取导致的准确性不足问题。提出一种基于深度迁移学习的快速辨识方法,通过分布式网络爬虫捕获数据,结合TF-IDF和近邻算法进行数据聚类和敏感信息提取。采用BERT-BiLSTM-CRF框架,融合深度迁移学习和特征融合策略,提取深层特征以实现快速准确辨识。实验结果显示,该方法在Kappa系数和辨识准确率上优于对比方法,有效提升了网络安全防护和用户隐私保障水平。This study aims to address the issue of insufficient accuracy in identifying online sensitive information caused by single-feature extraction in traditional methods.A rapid identification method based on deep transfer learning is proposed,which captures data through distributed web crawlers and combines TF-IDF and nearest neighbor algorithms for data clustering and sensitive information extraction.A BERT-BiLSTM-CRF framework is employed,integrating deep transfer learning and feature fusion strategies to extract deep-level features for fast and accurate identification.Experimental results show that this method outperforms comparative methods in terms of Kappa coefficient and identification accuracy,effectively enhancing network security protection and user privacy assurance.

关 键 词：深度迁移学习 网络敏感信息 特征提取 辨识模型 快速辨识 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]