Adblock Plus - free ad blocker

AI Security Review

Evidence context: threat category none; evidence quality strong.

Adblock Plus is among the most installed browser extensions anywhere, published at version 4.46.0 under the developer contact [email protected] with roughly 40 million users. This build declares no permissions and no host permissions, so whatever reach it has comes from what background.js does at runtime rather than from a broad manifest grant.

The headline number is 245,552 indicator entries, and it is the least useful thing in the data. The endpoint list runs 0-02.net, 0-100.it, 0-15.co.il, 000491b06a.com, 000scripps.hb.omtrdc.net, and continues for thousands more. That is a filter list. An ad blocker ships blocklists, and every domain it intends to block is a string inside its own bundle. Counting those strings as destinations the extension contacts inverts their purpose.

Three obfuscation findings deserve individual treatment. The high severity one, OBFUSCATION-INVISIBLE_ZERO_WIDTH-_locales/fa/messages.json-0, clears itself: that file is the Persian translation, and Persian writing uses zero-width non-joiners as ordinary letters. OBFUSCATION-LARGE_BASE64-background.js-39728 and OBFUSCATION-FUNCTION_INDIRECT-background.js-23453 sit inside a background.js that runs past 42,000 lines, which is the shape of bundled, machine-generated JavaScript. Holding a function in a variable and calling it indirectly is standard bundler output.

The network findings are the extension's own plumbing. NET-SOCKET_IO- and NET-FETCH- and NET-XMLHTTPREQUEST- entries in background.js and issue-reporter.js line up with filter updates and the broken-site report feature. issue-reporter.js exists to send a page report back to the vendor. Fetch and XHR calls inside a reporter script are the feature doing its job. Alongside them sit 1,488 code-smell matches and a single manifest-analysis entry, ordinary volume for a codebase this size.

A skeptic would point at 247,071 total findings and ask how a clean extension produces that many. The answer is arithmetic. Put a quarter-million domains in an embedded blocklist, run them through extractors that also catch property-access chains and IPv6 fragments, and that number falls out on its own. What matters is the buckets that are empty. No malware signatures, no malware findings, no exposed secrets, no tool-poisoning entries. The single high-severity finding has a locale file for a path. An extension smuggling a payload would leave matches in those buckets, and none of them has one. The only outbound path the code shows is the vendor reporting channel in issue-reporter.js. Nothing in this evidence describes behavior that departs from what Adblock Plus advertises.

Key Reasons

• Publisher is the verified Adblock Plus developer contact [email protected] at version 4.46.0 with about 40 million users.
• The 245,552 indicator entries are the extension's embedded blocklist; endpoints such as 0-02.net and 0-100.it are domains it blocks, not contacts.
• The only high-severity obfuscation finding is zero-width characters in _locales/fa/messages.json, ordinary Persian orthography.
• Zero malware signatures, zero exposed secrets and zero tool-poisoning findings across 247,071 total findings.
• Network findings in background.js and issue-reporter.js match filter updates and the vendor's broken-site report feature.

False Positive Considerations

• Ad blocker blocklist embedded in the bundle, producing 245,552 IoC entries that are block targets
• Zero-width characters in the Persian locale file _locales/fa/messages.json flagged as steganography
• Bundled/minified background.js triggering LARGE_BASE64 and FUNCTION_INDIRECT obfuscation rules
• Code-smell volume (1,488 matches) from a large bundled codebase

Reviewed 2026-10-15; recommended action: suppress false positive; model confidence 95%.