Measuring Risks to Users' Health Privacy Posed by Third-Party Web Tracking and Targeted Advertising

Research Background and Issues

• What problems or challenges did the authors identify?

  • Online advertising platforms may infer users' health conditions, potentially leading to risks surrounding health privacy, including unsolicited health ads, privacy breaches, and risks of social discrimination.
  • Existing studies have revealed the widespread use of third-party tracking technologies but lack empirical validation of whether advertising platforms can infer users' health conditions based on their browsing information.
  • Deceptive advertising techniques used in online health ads (e.g., pseudoscientific terms or fake customer reviews) may cause material or psychological harm to consumers.

• Why is this issue important?

  • Health information is highly sensitive, and its leakage or misuse may expose users to social stigma, price discrimination (e.g., increased insurance premiums), or even psychological trauma.
  • Deceptive health advertisements may mislead consumers, encouraging the use of ineffective or dangerous products, particularly among vulnerable groups or users with poor health conditions.

• Research Motivation and Related Work

  • This study complements existing literature by empirically analyzing whether advertising platforms target health-related ads based on users' browsing habits.
  • It expands the understanding of deceptive advertising techniques in online health ads and proposes policy recommendations to protect users' health privacy.

Solutions

• What methods or solutions did the authors propose?

  • Implemented an experimental data collection and analysis framework to measure the relationship between users' health-related web browsing behavior, online tracking, and ad content.
  • Conducted large-scale experiments using an improved Adscraper measurement framework to simulate user browsing histories and collect ad content.

• What are the innovative aspects of this solution?

  • By scraping actual user browsing histories and constructing reproducible simulated browsing profiles, the study controls for contextual variations in ad content and distinguishes between behavioral targeting and contextual targeting.
  • Combined qualitative and quantitative methods to analyze deceptive techniques in health-related ads and their distribution.

• What are the implementation steps and key technologies used?

  1. Data Collection Phase:
     • Recruited users and obtained 90 days of browsing history, distinguishing between health-related and non-health-related groups.
     • Automatically categorized and extracted features from the web pages visited by users and the ad content displayed.
  2. Experimental Phase:
     • Replayed users' browsing histories to simulate user behavior.
     • Used web crawlers to consistently visit 400 web pages and collect ads, distinguishing between behaviorally targeted and contextually targeted ads.
  3. Ad Classification and Deceptive Analysis:
     • Classified ad content using machine learning models and manually verified health-related ads.
     • Conducted qualitative analysis of deceptive techniques in ads, such as pseudoscience and fake reviews.
  4. Policy and Tool Improvements:
     • Enhanced Adscraper to support distributed parallel crawling, providing a technical foundation for future large-scale studies.

Research Findings

• What specific findings were achieved?

  • Users who viewed more health-related web pages were shown more health-related ads: browsing 100 health-related pages resulted in an increase of 2.3 health-related ads.
  • Among the collected health-related ads, 49.5% employed deceptive advertising techniques, such as pseudoscience, fake reviews, or manipulative content.
  • 70% of health-related websites embedded third-party trackers, with Google's trackers observing the majority of health-related web visits.
  • Ads targeted based on health conditions did not significantly increase, possibly due to limitations in ad runtime or data distribution.

• What advantages does it have compared to existing solutions?

  • Data collection was based on real users' browsing histories, complementing studies that rely solely on pre-configured profiles.
  • The dual focus on health privacy and deceptive advertising provides a more comprehensive understanding of the risks in the online advertising ecosystem.

• What were the experimental or evaluation results?

  • Quantitative regression analysis revealed a correlation between users' health-related browsing records and health-related ads.
  • Qualitative coding showed that categories such as supplements and medical devices had the highest proportion of deceptive ads.
  • Certain platforms, such as Taboola and Outbrain, raised concerns due to their high proportion of deceptive ads.

• Limitations and Future Directions

  • The data may not fully reflect behavioral characteristics during health crises (e.g., only 90 days of history).
  • Web crawlers may not fully replicate users' actual browsing experiences due to the lack of login states.
  • Future research could involve more real user participation or extend to social media ad ecosystems to enhance the scope of validation.
  • Stricter privacy and advertising regulations are recommended to limit tracking and the spread of misleading ads.

Conclusion

This study reveals preliminary evidence of the relationship between users' health-related browsing behavior and ad targeting, as well as the widespread use of deceptive techniques in online health ads. Through the improved Adscraper and detailed analysis of deceptive ads, this research advances the understanding of health privacy and online advertising regulation, contributing to policy discussions and recommendations.