Privacy-Friendly ID-Free Digital Ad Targeting Using URL Embeddings

This document is a peer-reviewed publication presented at the 2020 IEEE International Conference on Big Data, focusing on a novel approach to digital advertising that respects user privacy. It introduces ID-Free targeting, a method that enables effective ad targeting without relying on user profiles or historical data. The authors detail a two-step method: learning URL embeddings and training ID-free models. The paper outlines the challenges posed by the phasing out of third-party cookies and the need for privacy-conscious advertising strategies. The methodology involves using URL embeddings to predict user behavior based solely on the URL of the ad opportunity, thus facilitating targeted advertising without identifiable user information. The document also discusses the empirical results demonstrating the performance of the ID-Free model, emphasizing its ability to function effectively even with limited training data. The findings contribute to the ongoing discourse on privacy in digital advertising and provide insights into future methodologies that do not depend on traditional user tracking mechanisms.