Bluestreak — Privacy-Aware User Segmentation for Online Advertisement using Logistic Regression

Changelog for v1.0.1 (2025-07-15):

• Corrected cover title: “Linear Regression” → “Logistic Regression” and added hyphenation to “Privacy-Aware”
• Added PDF metadata (pdftitle, pdfauthor, pdfsubject)
• Minor typo and wording fixes

---

Abstract

The growing awareness of privacy in the digital world has not only made the block-

ing of third-party cookies more common but also introduced major regulatory changes

through the new European General Data Protection Regulation (GDPR). This regula-

tion has inherently changed the Internet in general and the online advertising industry in

particular: under these conditions, the traditional approach of tracking via user profiles

is becoming increasingly difficult. In this thesis, an alternative approach for predicting

age and gender segments of a user is proposed. With the presented Bluestreak method,

the sensitive data remains on the user’s device and only the anonymous segment pre-

dictions are sent back to the server. It differs from common approaches in that the

collection of the required data and the prediction of the desired segments is shifted to

the user’s browser. This approach is independent of tracking cookies and thus preserves

the user’s privacy. We conducted an evaluation on a real-world data set and show that it

is possible to improve the prediction accuracy for age and gender segments compared

to a User-Agent-based approach while only posing a low overhead on user’s devices.