Yuji Itadori Inner Collapse JJK Live Wallpaper

AI Security Review

Evidence context: threat category none; evidence quality moderate.

Yuji Itadori Inner Collapse JJK Live Wallpaper replaces the new tab page with Jujutsu Kaisen artwork. The manifest declares no permissions and no host permissions. That matters more than the finding count: without host permissions the extension cannot read cookies, tabs, history, or page content, and a background fetch to an arbitrary third-party host would be blocked unless that host allows cross-origin requests.

The behavior-level findings are five fetch calls: NET-FETCH-js/background.js-51, NET-FETCH-js/background.js-65, NET-FETCH-js/popup.js-39, NET-FETCH-js/sidePanel.js-362, and NET-FETCH-js/weather.js-49. Those line up with the product. A live wallpaper has to load images from somewhere, and js/weather.js points to a forecast widget on the new tab. There is no fetch tied to a login page, a search endpoint, or a cookie jar.

The 138 indicator entries are the main source of alarm, and they do not survive inspection. The listed hosts are two- and three-character strings: au.jo, fs.cy, 7k.sy, 9q.sx, bg.mp, ef.pr, el0.us, ff1.mx. One entry, cڽ.kg, has a non-ASCII character in the hostname. These have the shape of minified JavaScript tokens, variable.property chains and hex fragments, that the extractor re-reads as domains. They do not have the shape of a wallpaper CDN, a weather API, or an analytics service. No host in the list matches a known provider, and none is a search domain of the kind browser hijackers use.

The 43 code-smell matches are low severity and fire on ordinary JavaScript. No malware signature matched, no obfuscation was detected, and no hardcoded secret was found. For a new tab extension, the hijack risk is usually a search provider swap or a redirect through a monetized domain. Neither is visible here, and the empty permission set leaves little room for either.

A skeptic would say new tab wallpaper extensions are a known vehicle for search hijacking, and a domain list containing .su and .sx entries looks like a redirect chain. That concern is reasonable as a category. It needs a mechanism, and the mechanism is missing: no search new tab override beyond the wallpaper, no host permissions, no content script rewriting results, no obfuscated loader. The domains that would carry such a redirect are the same two-letter fragments that look like extraction noise, including the non-ASCII hostname. The publisher is a bare Gmail address and the listing reports no users, so there is no reputation to lean on; the verdict rests on the manifest and the nature of the code findings.

What would change this: a manifest listing host permissions or a search provider, a fetch to a named search or redirect domain, or any obfuscated payload in js/background.js. None of those are present in this version. The evidence describes a wallpaper extension making ordinary fetch calls, with a finding count built from extractor artifacts and code-quality rules.

Key Reasons

• Manifest declares no permissions and no host permissions, blocking cookie, tab, history and page-content access.
• No malware signature, obfuscation, secret or dependency finding matched in any file.
• The 138 indicator entries are two-to-three-character host fragments such as au.jo and fs.cy, plus the non-ASCII cڽ.kg, matching the known minified-token extraction pattern.
• The five network findings in js/background.js, js/popup.js, js/sidePanel.js and js/weather.js are consistent with loading wallpapers and a weather widget.
• No search provider swap, redirect host, or monetized new tab endpoint is visible.

False Positive Considerations

• IoC extractor misreading minified JavaScript tokens as domains, including two-letter hosts like au.jo and 7k.sy
• Low-severity code-smell rules firing on ordinary bundled JavaScript (43 matches)
• Network fetch findings flagging normal wallpaper and weather loading in background.js and weather.js
• A non-ASCII hostname, cڽ.kg, that cannot be a real service endpoint and points to extraction noise

Reviewed 2026-10-18; recommended action: suppress false positive; model confidence 82%.