Categories: Cyber Security News

Malicious Chrome AI Extensions Target 260,000 Users via Injected iFrames

As AI assistants like ChatGPT, Claude, Gemini, and Grok surge in popularity, cybercriminals exploit this trend through 30 fake Chrome extensions posing as productivity tools.

Security firm LayerX researchers uncovered the “Aiframe” campaign, impacting over 260,000 users by injecting remote-controlled iframes for surveillance and data theft.

These extensions, many featured in the Chrome Web Store, mimic AI summarizers, chat sidebars, translators, and Gmail helpers.

They share identical code, excessive permissions, and backend ties to tapnetic.pro. Instead of local processing, they overlay full-screen iframes from attacker subdomains like claude.tapnetic.pro, granting remote servers access to browser APIs without store updates or alerts.

The attack extracts tab content via Mozilla’s Readability library titles, text, and

metadata from any site, including authenticated pages. Voice data collection uses Web Speech API. Gmail-focused variants (15 total) inject scripts at document start on mail.google.com, persisting via MutationObserver to scrape emails and drafts, bypassing Google’s protections.

Featured “Claude” Assistant

Extension spraying evades takedowns: When “Gemini AI Sidebar” (ID: fppbiomdkfbhgjjdmojlogeceejinadg) was removed on February 6, 2025, it reappeared as “AI Sidebar” (gghdfkafnhfpaooiolhncejnlgglhkhe) on February 20 with unchanged malicious traits.

No direct CVEs link to these extensions, but they exploit Chrome’s extension model flaws. Here’s a summary table of the top affected extensions:

Sponsored
Extension ID Name Installs Risk Level (Est. CVSS equiv.) Description
fppbiomdkfbhgjjdmojlogeceejinadg Gemini AI Sidebar 80,000 8.8 (High) Iframe injection, Gmail scraping
nlhpidbjmmffhoogcennoiopekbiglbp AI Assistant 50,000 8.6 (High) Content extraction, voice recog
gghdfkafnhfpaooiolhncejnlgglhkhe AI Sidebar 50,000 8.8 (High) Republished variant, evasion
acaeafediijmccnjlokgcdiojiljfpbe ChatGPT Translate 30,000 7.5 (High) Tab metadata theft

Tapnetic.pro and onlineapp.pro host C2 subdomains themed to AI brands, masking as legit sites. IOCs include:

Indicator Type Value
C2 Domain tapnetic[.]pro
C2 Domain onlineapp[.]pro
Subdomain claude.tapnetic.pro
Subdomain chatgpt.tapnetic.pro
Subdomain gemini.tapnetic.pro

MITRE ATT&CK mappings: T1583 (Acquire Infrastructure), T1189 (Drive-by Compromise), T1036 (Masquerading), T1557 (Adversary-in-the-Middle), T1071.001 (Web Protocols C2).

Users should audit extensions, revoke permissions, and monitor for anomalies. Enterprises: Enforce allowlists. Expect more AI-themed threats as adoption grows.

Follow us on Google News , LinkedIn and X to Get More Instant UpdatesSet Cyberpress as a Preferred Source in Google

The post Malicious Chrome AI Extensions Target 260,000 Users via Injected iFrames appeared first on Cyber Security News.

rssfeeds-admin

Recent Posts

Preorders for Apple’s colorful MacBook Neo come with a $25 gift card

Last time I checked, the color indigo looked a little different. | Photo: Antonio G.…

4 minutes ago

Prediction markets in the news are a dangerous gamble

Today on Decoder, let’s talk about prediction markets, which continue to insert themselves into the…

4 minutes ago

RedAlert Mobile Espionage Campaign Targets Civilians with Trojanized Rocket Alert App for Surveillance

War zones have always been hunting grounds for opportunistic attackers, but the RedAlert mobile espionage…

18 minutes ago

RedAlert Mobile Espionage Campaign Targets Civilians with Trojanized Rocket Alert App for Surveillance

War zones have always been hunting grounds for opportunistic attackers, but the RedAlert mobile espionage…

18 minutes ago

RedAlert Mobile Espionage Campaign Targets Civilians with Trojanized Rocket Alert App for Surveillance

War zones have always been hunting grounds for opportunistic attackers, but the RedAlert mobile espionage…

18 minutes ago

New MongoDB Vulnerability Lets Hackers Crash Any MongoDB Server

A high-severity vulnerability, CVE-2026-25611 (CVSS 7.5), has been discovered in MongoDB, allowing unauthenticated attackers to…

18 minutes ago

This website uses cookies.