Microsoft Analyzes Chromium Extension That Intercepted Search Queries

· Original article ↗

Summary

Microsoft says a malicious Chromium extension impersonating Perplexity AI routed Omnibox searches and typed suggestions through attacker infrastructure, where its server code logged requests and headers. Google removed it after Microsoft's report; credential theft was未t

Key points

  • Microsoft identified a malicious Chromium extension impersonating Perplexity AI and setting itself as the browser's default search provider.
  • The extension routed queries and real-time typed suggestions through perplexity-ai[.]online before redirecting users to expected search results.
  • It used Manifest V3 settings overrides and declarativeNetRequest rules to intercept and redirect search traffic.
  • Microsoft found that shipped server code logged incoming requests and full HTTP headers; observed data included queries, user-agent information, and source IP addresses.
  • Microsoft said its analysis did not definitively confirm additional objectives such as credential theft.
  • Microsoft reported the extension to Google and said it had been taken down; recommended measures include allow-listing extensions and monitoring search-related network activity.

Article Details

Attack Vectors
  • A malicious Chromium extension impersonated Perplexity AI with similar branding and the typosquatted domain perplexity-ai[.]online.
  • The extension used chrome_settings_overrides to make its search provider the browser default and route Omnibox searches through attacker-controlled infrastructure.
  • Its suggest_url sent real-time search suggestions, including typed characters before submission, to the attacker-controlled domain.
  • The attacker server logged queries and full HTTP headers, including user-agent and source IP information, before declarativeNetRequest rules redirected users to search providers.
  • After installation, the extension opened an onboarding page at extension.tilda[.]ws/perplexityai.
Defensive Notes
  • Microsoft recommends allow-listing browser extensions and applying enterprise policy controls to restrict untrusted extensions.
  • Advise users to verify extension publishers, domains, and branding, especially for AI-themed tools.
  • Monitor browser search-setting changes, unusual extension permissions, and outbound search-related traffic to intermediary or non-standard domains.
  • Review extensions requesting search overrides or network-related APIs, and correlate indicators with behavioral signals and threat intelligence.
  • Microsoft also recommends platform protections, including Microsoft Edge extension protections and Microsoft Defender SmartScreen URL and web-content reputation checks.
  • The article reports no definitive evidence of credential theft; it identifies query and browsing-signal collection as observed risks.

Indicators of compromise

TypeIndicatorContext
DOMAINperplexity-ai[.]onlineTyposquatted, attacker-controlled domain used to intercept search traffic and collect queries.
URLhxxps[:]//extension[.]tilda[.]ws/perplexityaiSpecific onboarding page opened by the malicious extension after installation.
URLhxxps[:]//perplexity-ai[.]online/search/{searchTerms}Attacker-controlled search endpoint that receives queries before redirecting users to search providers.

MITRE ATT&CK

Vendors

Products

Related Articles