There is a newer version of the record available.

Published September 18, 2026 | Version v1
Other Restricted

One App, Many Realities: A Cross-Store Study of Security and Privacy Disclosures in XR Apps (AEC-version)

Authors/Creators

  • 1. ROR icon University of California, Irvine
  • 2. ROR icon George Mason University

Description

Artifact for the paper, "One App, Many Realities: A Cross-Store Study of Security and Privacy Disclosures in XR Apps" (NDSS 2027). The study examines cross-store consistency in safety, security, and privacy metadata for 9,574 XR app listings (1,063 apps) across 7 storefronts and analyzes 1.26M user reviews to assess whether they reveal inconsistencies.
 
Availability
The AE-specific archive, including the database, models, and APKs, is not shared via a public link. We do not publicly release it because the metadata inconsistencies could enable unintended or inappropriate use. Moreover, given the scale of the dataset, individually notifying every developer whose application may exhibit an inconsistency was not feasible.

For availability, we have published the code itself (without the underlying dataset, model checkpoints, and app binaries) at https://doi.org/10.5281/zenodo.22764919. To obtain results, users of the public release can either contact the authors for the full dataset (noted in that code's README and in the paper) or populate the skeleton database (data/xr_inconsistencies_skeleton.db) with their own collected and normalized XR listings by following data/SCHEMA.md and the mapping tables in the paper's appendix. Since the schema is not tied to specific stores or apps, the same scripts can also analyze apps from other XR stores with similar fields.

AE Archives
The complete artifact is 20+ GB, so it is split into zips due to size and Zenodo's upload limits. Extract every one downloaded into the same target folder (they share the same internal paths). Not all archives are required.
  • ndss_ae_core.zip (required, ~3GB download / ~6GB unpacked): code, full DB, the 2 review-classifier checkpoints behind the paper's reported results, and the static-analysis evidence. Enough to run the default pipeline and reproduce claims except the RoBERTa-vs-DeBERTa comparison tables and the complete static analysis consistency checks (Androguard-based, needs raw APKs).
  • ndss_ae_static_apks_{meta,pico}_part{1,2}.zip (four optional zips, ~10GB download / ~11GB unpacked): raw APKs. Only needed to regenerate the static-analysis evidence (decompile/dump from scratch) and the complete consistency check (with Androguard). Download all four if doing this step.
  • ndss_ae_checkpoints_comparison.zip (optional, ~5GB download / ~6GB unpacked): the other 4 review-classifier checkpoints. Only needed for the RoBERTa-vs-DeBERTa / BERT-vs-LLM comparison tables.
Disk space: ~6GB with core alone. ~23GB unpacked for all three.
 
See the README for more details.

Files

Restricted

The record is publicly accessible, but files are restricted. Log in to check if you have access.