Scholarlink

<p>A privacy policy is a legal document that explains how websites and mobile applications collect, store, and share user information. However, these documents are often lengthy and complex, leading to user confusion and limited understanding. To address this issue, researchers have explored Machine Learning (ML) and deep learning (DL) techniques to enhance the transparency of mobile app privacy policies. This study provides a comprehensive review of existing literature, highlighting commonly used datasets, applied methodologies, and current research gaps. Key datasets analyzed include OPP-115, APP-350, Google Play Store App Reviews, and the Mobile App Reviews Summarization (MARS) Dataset, which offer annotated examples of privacy regulations, app descriptions, and user feedback to support model training and evaluation. The review identifies several effective approaches, such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and large language models (LLMs) like GPT-4 and ViT, for extracting features and classifying privacy-related content. Despite these advancements, challenges remain regarding annotation quality, dataset imbalance, and comprehensive coverage of privacy concerns. The findings emphasize the importance of transparency in privacy policies and the role of ML and DL in improving user understanding. Furthermore, the review highlights the need for large-scale, balanced datasets, innovative model architectures, and robust evaluation metrics. It also underscores the importance of integrating user feedback mechanisms into privacy policy analysis frameworks to foster continuous improvement and promote trust and accountability in digital environments. The review further reveals a critical transparency gap in the current literature. While ML and DL models demonstrate strong performance in detecting privacy-related disclosures within mobile app policies, they remain limited in evaluating the legal adequacy, specificity, and user-centric clarity of these disclosures. This highlights the need for future research to move beyond detection-based approaches toward frameworks capable of assessing meaningful transparency in accordance with regulatory requirements such as GDPR and POPIA.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Recent advances in machine learning and deep learning methodologies for responsible artificial intelligence and mobile application privacy transparency

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月24日

查看全文
  • Samantha Danster,
  • Andronicus A. Akinyelu,
  • Micheal O. Olusanya,
  • Mokotsolane Mase

摘要

A privacy policy is a legal document that explains how websites and mobile applications collect, store, and share user information. However, these documents are often lengthy and complex, leading to user confusion and limited understanding. To address this issue, researchers have explored Machine Learning (ML) and deep learning (DL) techniques to enhance the transparency of mobile app privacy policies. This study provides a comprehensive review of existing literature, highlighting commonly used datasets, applied methodologies, and current research gaps. Key datasets analyzed include OPP-115, APP-350, Google Play Store App Reviews, and the Mobile App Reviews Summarization (MARS) Dataset, which offer annotated examples of privacy regulations, app descriptions, and user feedback to support model training and evaluation. The review identifies several effective approaches, such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and large language models (LLMs) like GPT-4 and ViT, for extracting features and classifying privacy-related content. Despite these advancements, challenges remain regarding annotation quality, dataset imbalance, and comprehensive coverage of privacy concerns. The findings emphasize the importance of transparency in privacy policies and the role of ML and DL in improving user understanding. Furthermore, the review highlights the need for large-scale, balanced datasets, innovative model architectures, and robust evaluation metrics. It also underscores the importance of integrating user feedback mechanisms into privacy policy analysis frameworks to foster continuous improvement and promote trust and accountability in digital environments. The review further reveals a critical transparency gap in the current literature. While ML and DL models demonstrate strong performance in detecting privacy-related disclosures within mobile app policies, they remain limited in evaluating the legal adequacy, specificity, and user-centric clarity of these disclosures. This highlights the need for future research to move beyond detection-based approaches toward frameworks capable of assessing meaningful transparency in accordance with regulatory requirements such as GDPR and POPIA.