融合敏感性与关联性的隐私风险评估及实证研究

打开文本图片集

关键词：敏感性；关联性；隐私风险评估；在线社区；小木虫；BERT-BiLSTM-CRF DOI：10.3969/j.issn.1008-0821.2026.04.012 [中图分类号]G203；TP309 ［文献标识码]A[文章编号】1008-0821（2026）04-0136-13

Abstract：[Purpose/Significance]This study aims to evaluate the privacy risksasociated with unstructured text within online communityuser-generated content，adressingthechallnge thatsuch data poses asahigh-riskvector for privacyleakagedue to contextual variations in sensitivityand corelation-based vulnerabilities.[Method/Process]Aprivacyriskquantification framework integrating sensitivityandrelevancewasproposed，with experimentalvalidatioconducted on the‘Member Networking’section of theacademic platform Muchong.This framework employs a BERTBiLSTM-CRFdeeplearning model toachieve atribute extraction fromunstructuredtext.Atributesensitivitywas quantifiedusing aprivacylexicon，atributecorrelation was measuredvia Pointwise Mutual Information（PMI），and thesefactors were integratedwith privacyprincipal identification metrics tocomputeprivacyriskscores，folowed byrisk stratification. [Result/Conclusion]Ablationstudiesandmanualvalidationdemonstrate itscapabilitytoidentify，sess，and tratifyprivacyrisks inunstructured textualdata.These findings ofer new insights forimproving privacy protectionpolicies and platform privacy governance.

Keywords：sensitivity；correlation；privacy risk assessment;online community;Muchong；BERT-BiLSTM-CRF

在线社区是人们进行信息交流、传播与共享的重要场域。（剩余19656字）