The Risks of Local Learning: Understanding and Attacking Safari’s Privacy Features

Artifact for our The Risks of Local Learning: Understanding and Attacking Safari’s Privacy Features paper to be published in Proceedings on Privacy Enhancing Technologies 2027 (Issue 1).

Prerequisites

• Docker and Docker Compose installed (consider adding yourself to the docker group see more)

Global Setup (Dataset)

Create and download a shared volume for the dataset:

docker volume create zenodo-data
docker build -t zenodo-fetcher ./dataset
docker run --rm -v zenodo-data:/data zenodo-fetcher

Components

This section describes the different components we designed for this paper, structured as different subdirectories in this repository. For detailed information, refer to the README.md in each subdirectory.

Crawling Infrastructure

The crawler directory contains the crawling infrastructure used for our Tranco Top 10K crawls.

Data Donation Platform

The data-donation directory contains the data donation platform used for our user study. It consists of the study web application and the analytic scripts used to analyze the collected data.

ITP Circumvention Analysis

The itp-circumvention directory contains the scripts used to analyze ITP circumvention in the wild.

Proof-of-Concept Code

The poc directory contains the proof-of-concept code used to demonstrate the attacks described in our paper.

WebPrivacy List Evaluation

The webprivacy-blocking-performance directory contains the scripts used to evaluate the performance of the WebPrivacy blocking lists in comparison to the reference lists.

WebPrivacy List Crawler

The webprivacy-crawler directory contains the WebPrivacy list crawler used to periodically fetch all WebPrivacy lists.

WebKit Patches

The webkit-patches directory contains the WebKit patches for enhanced logging.