Fake “AI Tool” Ads Drop MediaArena Persistence

Heimdal Security · Medium sophistication
Last updated July 30, 2026

A malvertising campaign is luring users with fake free “AI tool” downloads (recipe/meal-planning apps) delivered via paid search ads. Even when Microsoft Defender later quarantines the detected file, the installer can already have created persistence (Startup shortcut and HKCU Uninstall key), meaning the device may still be compromised. The article provides specific domains, hosting infrastructure, and the exact persistence checks defenders should perform.

How the attack worked

This campaign relies on paid search ads to promote fake free Windows "AI tools," currently themed as recipe and meal-planning apps. Users searching for free software click a paid result, land on a legitimate-looking site, and download an installer that does not require admin rights. Once run, the installer writes files to AppData, adds a Start Menu shortcut, creates an HKCU Uninstall registry key, and drops a Startup folder shortcut for boot persistence. The loader itself, referred to as math.dll, is injected in memory rather than dropped to disk, which limits what file-based scanning alone can catch.

Why it succeeded

The campaign's success comes down to timing and rotation. Persistence was observed being created around 21 seconds into execution, while Microsoft Defender's quarantine action occurred around 29 seconds later. That gap means the malicious artifacts can already be embedded on the endpoint by the time an alert fires. On top of that, the operators rotate lure brand names and domains quickly, for example GiveMeRecipe, KitchenCanvas, and FoodFormula, all fronting the same underlying toolkit. This rotation means indicators like brand names or domains have a short useful life for detection purposes.

What to watch for

  • Paid search ads offering a free "AI tool" tied to an unfamiliar or off-brand app name
  • Recently created or rotating domains hosting the download
  • An installer prompting for a vague AI-related benefit rather than a specific, verifiable feature
  • A Startup folder shortcut tied to the downloaded app's name
  • An HKCU Uninstall registry key mimicking a legitimate software install

Any of these persistence artifacts on a host that triggered a MediaArena-related alert should be treated as a sign the device may still be compromised, even if the file itself was quarantined.

How to build resistance

Defenders should treat a quarantine alert as the start of an investigation rather than the end of one. Verifying whether anything executed and whether persistence artifacts exist is a more reliable process than relying on the quarantine result alone. Because brand names and domains rotate quickly, detection strategies should lean on behavior-based signals and known persistence mechanisms, such as Startup folder shortcuts and HKCU Uninstall keys, rather than static indicators. End users across all departments, not just technical staff, should be made aware that free "AI tool" ads served through paid search are a common lure format, so they think twice before downloading unfamiliar software promoted this way. SOC and incident response teams should incorporate these specific artifact checks into their playbooks for any related alert.

Key findings

  • Microsoft Defender quarantine may occur after persistence is already written (persistence observed at 21 seconds; quarantine at 29 seconds).
  • Campaign uses paid search ads to push fake free Windows “AI tool” apps themed as recipe/meal-planning tools.
  • Installer does not require admin rights and writes to AppData, adds Start Menu shortcut, creates an HKCU Uninstall key, and creates a Startup folder shortcut for boot persistence.
  • Infrastructure and branding rotate quickly (multiple lure brands/domains), so behavior/persistence artifacts are more reliable than brand-string indicators.
  • The loader (math.dll) is injected in memory rather than dropped to disk, so defenders should hunt for persistence artifacts instead of the file.

Who’s being targeted

  • Commonly targeted roles: All employees, End users who install software, IT support/helpdesk, SOC/incident response.
  • Affected industries: Multiple/various industries (more than 20 client environments).
  • Attack channels: website.
  • Impersonated: A legitimate-looking free app brand (rotating names such as GiveMeRecipe/KitchenCanvas/FoodFormula).

Red flags to watch for

  • Off-brand “free AI tool” promoted via paid ad rather than an official vendor site
  • Recently created/rotating domains and brand names
  • Installer asks to run a downloaded EXE for a vague “AI tool” benefit
Try Mirage

Mirage safely runs attacks like this one against your own team, so you find out what happens before a real adversary does.

Get a demo

Frequently asked questions

What is the MediaArena malvertising campaign?

It is a malvertising campaign that uses paid search ads to promote fake free Windows AI tools, currently themed as recipe and meal-planning apps, to trick users into downloading malicious installers.

Does a Microsoft Defender quarantine mean the threat is neutralized?

Not necessarily. Persistence has been observed being written around 21 seconds into execution, while quarantine occurred around 29 seconds, meaning the malicious artifacts can already be in place before detection.

What persistence artifacts should defenders check for?

Defenders should look for a Startup folder shortcut tied to the app name and an HKCU Uninstall registry key mimicking a legitimate install. If either is present, the host should be treated as still compromised.

Why do brand names and domains matter less than behavior in this campaign?

The campaign rotates lure brands and domains quickly, so any single brand or domain indicator has a short shelf life, making behavior and persistence artifacts more reliable signals to hunt on.

Read the video transcript

You search for a free AI meal-planning app, click a paid ad for “GiveMeRecipe,” and grab the Windows download. Looks legit, right? Behind that ad is a malvertising campaign. The fake AI app installer runs without admin rights, drops files into AppData, adds itself to your Start Menu, and quietly sets a Startup shortcut so it launches every boot. Here’s the trap: Microsoft Defender might pop up and quarantine MediaArena a few seconds later, but by then the damage is done. Persistence was created at around 21 seconds; quarantine hit at 29. The loader hides in memory, and the brand names, GiveMeRecipe, KitchenCanvas, FoodFormula, rotate constantly. So if you ever see MediaArena quarantined after installing a 'free AI' recipe tool from a paid ad, don’t assume you’re clean, open a ticket and tell IT to check your Startup folder and HKCU Uninstall keys for that app name.

Similar attacks

Korea Flags Job-Offer Phish + Watering Holes

Korea Flags Job-Offer Phish + Watering Holes

South Korean agencies warned that a state-backed hacking group is actively targeting citizens and businesses using job-themed phishing emails and “watering hole” attacks on legitimate websites. The phishing lures include fake job applicants sending resume links and impersonated recruiters sending…

July 31, 2026
Fake Claude App and Alert Apps Drive New Scams

Fake Claude App and Alert Apps Drive New Scams

This roundup describes multiple real-world campaigns where attackers trick people into installing malicious software that looks legitimate (a fake Claude desktop app, a fake emergency alert app, and banking-malware phishing). The common pattern is “looks normal, feels urgent,” leading users to…

July 23, 2026
ClickFix Trick Spreads ACR Stealer via Paste-Run

ClickFix Trick Spreads ACR Stealer via Paste-Run

Microsoft observed real-world campaigns where victims were tricked by “ClickFix” prompts into pasting a command into Windows Run, which then installed ACR (Amatera) Stealer. The malware steals saved browser passwords, live session tokens, and Microsoft 365/OneDrive/SharePoint files, meaning…

July 17, 2026
Fake CAPTCHA Tricks Users Into Running Malware

Fake CAPTCHA Tricks Users Into Running Malware

Ukraine’s CERT-UA says Russian-linked attackers are using fake CAPTCHA prompts on hacked websites to trick people into running malicious PowerShell commands themselves. The page pretends it’s verifying the visitor is human, but instead guides them to open the Windows Run box and execute…

July 21, 2026
Kimsuky Uses AI-Polished Phishing Lures

Kimsuky Uses AI-Polished Phishing Lures

Researchers say North Korea-linked Kimsuky is using AI tools to improve phishing campaigns that deliver malware through ZIP files containing malicious Windows shortcut (LNK) files. The lures are designed to look like legitimate international event materials, research reports, or meeting requests,…

August 10, 2026
Fake VS Code Extensions Spied on Devs and CI/CD

Fake VS Code Extensions Spied on Devs and CI/CD

Researchers found a short burst of counterfeit extensions on the Open VSX marketplace that copied the names and descriptions of legitimate Visual Studio Code extensions. Some of these “evil twin” extensions quietly sent developer and CI/CD environment details to a newly registered domain, creating…

August 10, 2026