AdGuard AdBlocker

AI Security Review

Evidence context: threat category none; evidence quality strong.

AdGuard AdBlocker is published by AdGuard ([email protected]), the company behind one of the most widely used ad-blocking tools. The extension declares no special permissions and no host permissions, which is appropriate for a filter-based ad blocker. The 494,959 findings appear alarming in raw count, but the composition tells the real story: zero malware signatures matched, despite exhaustive scanning.

The obfuscation findings are false positives from well-documented sources. The OBFUSCATION-INVISIBLE_ZERO_WIDTH detections in _locales/fa/messages.json and filters/filters_i18n.json are legitimate zero-width joiners and non-joiners used in Farsi text rendering. These are not steganography or payload obfuscation, but essential characters for Persian script. The OBFUSCATION-UNICODE_HEAVY findings in vendored files like vendors/tswebextension.js and vendors/tsurlfilter.js are minified JavaScript from bundled libraries, not malicious obfuscation.

The network findings (12 total) are generic FETCH and XMLHTTPREQUEST calls detected in shared editor modules (shared/editor.js), popup UI code (pages/popup.js), and xstate state-machine library (vendors/xstate.js). These are normal patterns for any extension with a UI and configuration system. No suspicious domains appear in the network_endpoints list provided; the endpoints are mostly placeholder or test domains used in blocklists themselves, not command-and-control servers.

The 494,098 IoC findings are the real driver of finding volume. Ad blockers embed thousands of domain names in their blocklists to filter ads and trackers. The file paths confirm this: ruleset_6.json, ruleset_11.json, and ruleset_235.json are declarative filter rules, not executable code. Each domain in those rulesets gets flagged as an IoC, creating the illusion of a security issue where none exists.

A skeptic might argue that high finding counts should still trigger caution, or that obfuscation anywhere is a warning sign. But in the context of a known publisher, zero malware signatures, and findings that are entirely explainable by ad-blocking architecture, the evidence points cleanly to a legitimate tool being flagged by the scanner's known false-positive patterns. AdGuard has 17 million users and is a mainstream extension; the absence of any malware signatures despite this scale is strong evidence that the findings are noise.

Key Reasons

• Developer identity is verified: [email protected] is the official AdGuard contact.
• Zero malware signatures matched despite 494K findings (malware-signature: 0).
• OBFUSCATION-INVISIBLE_ZERO_WIDTH findings are in i18n locale files (_locales/fa/messages.json, filters/filters_i18n.json), which use legitimate zero-width characters for Farsi and other scripts.
• Network findings are generic FETCH/XMLHTTPREQUEST calls in bundled libraries (xstate.js, tswebextension.js) and editor modules, not to suspicious domains.
• The 494K IoC count is noise from embedded blocklists (ruleset_6.json, ruleset_11.json, ruleset_235.json), expected for an ad blocker.

False Positive Considerations

• Bundled blocklists trigger massive IoC counts (494,098 IoCs from embedded domain lists).
• Zero-width Unicode characters in Farsi locale file are legitimate script support, not steganography.
• Minified vendored libraries (xstate.js, tswebextension.js) trigger generic network and obfuscation rules.
• Ad blockers inherently embed thousands of domains in their filtering rules.

Reviewed 2026-09-30; recommended action: suppress false positive; model confidence 92%.