Privacy Policy

1. Summary

Haram Blur is a Chrome extension that blurs adult photos on your device. A photo stays covered until its own scan finishes, then stays blurred when the highest enabled class (Porn, Hentai, or Sexy) meets your threshold. The model is included in the extension. We do not operate an account system, analytics service, image-processing server, or remote configuration service. Detection can miss content, and safe photos may be blurred. You can reveal a mistake. A Firefox add-on is not published yet.

2. What the extension accesses

To filter images, the extension may process the following information on your device:

• Website photos: image elements, enabled CSS background photos, and enabled video poster images on pages you visit.
• The current page address: used locally so you can pause or resume that site, and so the extension can update the open page. This is not a browsing-history record.
• Your settings: protection status, threshold, blur strength, enabled model categories, reveal preferences, and hosts you explicitly pause.

3. Local image processing

TensorFlow.js and the bundled NSFWJS model run inside a hidden offscreen extension page with canvas and WebGL. Photos are not sent to Haram Blur or a third-party AI API.

If a website prevents direct canvas access, the offscreen page may fetch an HTTP, HTTPS, or user-enabled local-file image from its source without cookies. The request times out after 8 seconds, rejects non-image responses and images larger than 2 MiB, resizes accepted images for local classification, and then discards the bytes.

4. What we do not collect

We do not collect names, email addresses, account identifiers, browsing history, photos, location, health or financial data, advertising identifiers, or usage analytics. We do not sell or transfer user data for advertising, credit, or unrelated purposes.

5. Storage and retention

Settings are stored in chrome.storage.local on your device. A limited in-memory cache keeps recent classification results for the current page so the same photo is not processed repeatedly. Page results disappear when the tab or page session ends. Settings remain until you reset them or uninstall the extension.

6. Permissions

• storage — saves your settings on this device.
• offscreen — runs the local model in a hidden page that can use canvas and WebGL.
• tabs — reads the active page address locally when you pause or resume a site, and messages open tabs when settings or the scanner need an update. This is not a browsing-history record.
• alarms — about every 30 minutes, checks on this device that local protection is still running.
• webNavigation — notices a new top-level http, https, or file page so scanning can start again. These events are not saved.
• notifications — shows a notice only when that local check finds protection unavailable or degraded.
• favicon — shows the current site icon beside the pause control. Firefox does not receive this permission.
• Host access (<all_urls>) — reads photos on the sites you open. Local files are included only after you enable Chrome’s “Allow access to file URLs”. That Chrome setting stays off until you turn it on.

7. Remote code and third parties

Haram Blur does not download or execute remote JavaScript. Its model and application code are packaged with the extension. NSFWJS and TensorFlow.js run locally and are subject to their respective open-source licenses.

8. Your choices

• Turn protection off from the popup.
• Pause filtering for a specific site.
• Reveal one blurred photo, or reveal and re-hide every blurred photo on the current page.
• Rescan the current page.
• Change which model categories count toward the threshold.
• Reopen the onboarding guide from Settings.
• Reset settings or uninstall the extension at any time.

9. Children

Haram Blur is a general content filter. It is not directed at children and does not attempt to identify anyone shown in a photo.

10. Changes

If this policy changes, the “Last updated” date above will be revised.

11. Contact

Questions about this policy can be sent to msaidzengin@gmail.com.