English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Behavioral data generated by users' devices, ranging from emoji use to pagesvisited, are collected at scale to improve apps and services. These data,however, contain fine-grained records and can reveal sensitive informationabout individual users. Local differential privacy has been used by companiesas a solution to collect data from users while preserving privacy. We herefirst introduce pool inference attacks, where an adversary has access to auser's obfuscated data, defines pools of objects, and exploits the user'spolarized behavior in multiple data collections to infer the user's preferredpool. Second, we instantiate this attack against Count Mean Sketch, a localdifferential privacy mechanism proposed by Apple and deployed in iOS and Mac OSdevices, using a Bayesian model. Using Apple's parameters for the privacy loss$\varepsilon$, we then consider two specific attacks: one in the emojis setting-- where an adversary aims at inferring a user's preferred skin tone for emojis-- and one against visited websites -- where an adversary wants to learn thepolitical orientation of a user from the news websites they visit. In bothcases, we show the attack to be much more effective than a random guess whenthe adversary collects enough data. We find that users with high polarizationand relevant interest are significantly more vulnerable, and we show that ourattack is well-calibrated, allowing the adversary to target such vulnerableusers. We finally validate our results for the emojis setting using user datafrom Twitter. Taken together, our results show that pool inference attacks area concern for data protected by local differential privacy mechanisms with alarge $\varepsilon$, emphasizing the need for additional technical safeguardsand the need for more research on how to apply local differential privacy formultiple collections.

Pool Inference Attacks on Local Differential Privacy: Quantifying the Privacy Guarantees of Apple's Count Mean Sketch in Practice