面向APP应用的隐私合规的检测方法
Privacy Compliance Detection Methods for Applications
何艾星 1郑旭飞 1谢明天 1何枘峰1
作者信息
- 1. 西南大学 计算机与信息科学学院/软件学院,重庆 400715
- 折叠
摘要
面对严峻的个人信息保护监管态势,针对如何科学评估 APP应用的隐私合规问题,提出了一种基于 BERT-TextCNN分类模型的隐私合规检测方法.该方法充分考虑国内现有的法律法规,构建了一套完整的隐私合规测评指标体系,通过人工标注,建立了专业的隐私合规语料库,并通过该语料库训练 BERT-TextCNN 分类模型,以精准提取出 APP应用隐私政策的合规要点.最终,采用该模型的预测结果进行隐私合规得分计算,以全面检测该APP应用可能存在的隐私合规风险点,保障其隐私政策的完整性.实验结果表明,该模型在高效、准确地提取隐私合规要点和检测 APP应用的合规风险点方面表现卓越.这一创新性方法有望为企业和开发者防范监管通报的风险提供指导,同时为监管部门提供有益的参考,对于保障个人隐私权益具有积极的意义.
Abstract
Facing a severe regulatory landscape on personal information protection,a method for scientifically assessing privacy compliance issues in applications(APPs)is proposed.This method utilized a BERT-TextCNN classification model for privacy compliance detection.Taking into full consideration of the existing domestic laws and regulations,a comprehensive set of priva-cy compliance evaluation indicators was established.Through manual annotation,a specialized corpus for privacy compliance was built.The BERT-TextCNN classification model was then trained on this corpus to accurately extract key points from the privacy policies of APPs.Ultimately,the model's predictions were used to calculate a privacy compliance score,providing a thorough examination for potential privacy compliance risks in the APPs.This approach ensures the integrity of the privacy pol-icies.Experimental results demonstrated the outstanding performance of the model in efficiently and accurately extracting priva-cy compliance points and detecting compliance risks in APPs.This innovative method holds the promise of guiding enterprises and developers in avoiding the risks associated with regulatory notifications.Simultaneously,it provides valuable references for regulatory authorities,contributing positively to safeguarding individuals'privacy rights.
关键词
应用程序/隐私保护技术/文本分类/深度学习/指标体系Key words
application program/privacy protection technology/text classification/deep learning/index system引用本文复制引用
基金项目
重庆市软件产品质量提升及测试标准化研究与应用技术创新与应用发展(面上项目)(cstc2020jscxmsxmX0148)
出版年
2024
NETL
NSTL
万方数据