HARPO: Learning to Subvert Online Behavioral Advertising

—Online behavioral advertising, and the associated tracking paraphernalia, poses a real privacy threat. Unfortu-nately, existing privacy-enhancing tools are not always effective against online advertising and tracking. We propose H ARPO , a principled learning-based approach to subvert online behav- ioral advertising through obfuscation. H ARPO uses reinforcement learning to adaptively interleave real page visits with fake pages to distort a tracker’s view of a user’s browsing profile. We evaluate H ARPO against real-world user profiling and ad targeting models used for online behavioral advertising. The results show that H ARPO improves privacy by triggering more than 40% incorrect interest segments and 6 × higher bid values. H ARPO outperforms existing obfuscation tools by as much as 16 × for the same overhead. H ARPO is also able to achieve better stealthiness to adversarial detection than existing obfuscation tools. H ARPO meaningfully advances the state-of-the-art in leveraging obfuscation to subvert online behavioral advertising.