Join our Newsletter — 33% off our NHI Course
Home› Guides› ClickFix and Fake CAPTCHA Paste-and-Run Attacks Guide: How…
Guide Attack Techniques & Malware

ClickFix and Fake CAPTCHA Paste-and-Run Attacks Guide: How the Lure Works and How to Stop It

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 14 min read
On this page

ClickFix attacks persuade a user to copy a command from a web page and run it themselves, so the malware arrives through the keyboard rather than as a file that email gateways and browsers can inspect. The lure is usually a fake CAPTCHA, a "verify you are human" box or a fake error with a "fix" button. ESET reported that ClickFix grew by over 500% in the first half of 2025 compared with the second half of 2024 and became the second most common attack vector in its telemetry after phishing, and it is now used by commodity infostealer operators, ransomware affiliates and state-sponsored groups. This guide explains the attack chain, where it shows up, how to detect it from registry, process and script telemetry, and which Windows and macOS settings remove most of the risk. For endpoint engineers, SOC analysts and awareness leads; it goes deeper than the Phishing Guide.

Key takeaways

  • The user is the execution engine. A script on the lure page writes a command to the clipboard and the page tells the victim to press Win+R (or open Terminal), paste and press Enter, which sidesteps controls built around downloaded files.
  • It is a mainstream technique, not a niche trick. MITRE ATT&CK tracks it as T1204.004 Malicious Copy and Paste, and CISA's Interlock advisory, Microsoft and Proofpoint all document its use by crime and state actors.
  • Removing the Run dialog where it is not needed is the single most effective control. Combine it with application control on mshta, PowerShell and other script hosts for standard users.
  • The RunMRU registry key is your best forensic artefact. Commands pasted into Run are recorded under the user's hive, and explorer.exe spawning powershell.exe or mshta.exe with an encoded command line is a strong detection.
  • Awareness must teach one rule: no legitimate website, CAPTCHA or IT fix will ever ask you to paste something into Run, Terminal or PowerShell.

How the attack works

ATT&CK describes Malicious Copy and Paste as adversaries relying on a user to copy and paste code into a command interpreter, and names ClickFix as the strategy of presenting fake fixes or CAPTCHAs that instead instruct the user to paste malicious code. The chain is short, which is why it is effective:

  1. Lure delivery: the victim reaches a lure page through a phishing link or HTML attachment, a malvertising redirect, or a compromised legitimate website that injects the fake overlay.
  2. Clipboard staging: clicking "I am not a robot" or "Fix it" runs JavaScript that silently places a command on the clipboard. The user never sees the text unless they paste it somewhere visible.
  3. Guided execution: the page shows numbered "verification steps": press Windows+R, press Ctrl+V, press Enter. Variants use the Win+X menu to open PowerShell or Windows Terminal, and macOS variants tell the user to open Terminal.
  4. Living-off-the-land download: the pasted line is typically an obfuscated or Base64-encoded call to powershell.exe, mshta.exe, curl or a similar built-in binary that fetches the next stage (T1059.001 PowerShell, T1218.005 Mshta, T1105 Ingress Tool Transfer).
  5. Payload and persistence: Microsoft reports that final payloads are often fileless and run in memory through binaries such as msbuild.exe, regasm.exe or powershell.exe, followed by Run keys, Startup folder entries or scheduled tasks.

Microsoft's analysis of a macOS campaign impersonating an internet provider shows the same pattern adapted for Apple devices: the pasted command repeatedly prompted for the system password, validated it, downloaded Atomic macOS Stealer and used the password to remove the quarantine attribute before running it. See the Atomic macOS Stealer (AMOS) Guide.

Why it bypasses traditional controls

ControlWhy ClickFix gets past itWhat still helps
Email attachment sandboxingNo malicious file is attached; the lure is a link or a harmless-looking HTML pageTime-of-click URL rechecks, blocking newly registered and typosquatted domains
Browser download protectionThe browser never downloads the payload; a native binary does after the pasteDNS and web filtering applied to the whole device, not just the browser
Mark of the Web and SmartScreen promptsContent fetched by PowerShell or curl in memory does not trigger the file reputation prompt in the same wayApplication control and attack surface reduction rules on script behaviour

Where it shows up

  • Infostealer distribution: Microsoft describes Lumma Stealer as the most prolific ClickFix payload, alongside Lampion and AMOS, and CISA notes ClickFix use in Lumma Stealer and DarkGate campaigns. See the Lumma Stealer Guide.
  • Remote access trojans and RMM tools: Microsoft lists Xworm, AsyncRAT, NetSupport, SectopRAT and ScreenConnect, and Unit 42 investigated NetSupport campaigns using DocuSign and Okta themed lures.
  • Loaders that hand off to ransomware: Latrodectus and MintsLoader are delivered this way, and the FBI observed the Interlock ransomware group using fake CAPTCHA pages to run Base64-encoded PowerShell. See the Interlock Ransomware Profile Guide.
  • State-sponsored espionage: Proofpoint documented ClickFix use between October 2024 and April 2025 by TA427 (North Korea, overlapping Kimsuky), TA450 (Iran, overlapping MuddyWater), TA422 (Russia, overlapping APT28, as reported by CERT-UA) and a suspected Russian cluster it tracks as UNK_RemoteRogue. Lures included fake diplomatic meeting requests and a fake reCAPTCHA.
  • Compromised legitimate sites: Microsoft reports Storm-0249 injecting fake Cloudflare Turnstile checks into real websites, and CISA calls drive-by delivery from compromised legitimate sites unusual for a ransomware group.

Impact

  • Credential and session theft: stealers take browser-saved passwords and cookies (T1555.003 Credentials from Web Browsers), which means valid sessions to SaaS, cloud consoles and VPNs that may bypass MFA. See the Infostealer Logs and Session Token Abuse Guide.
  • Developer and machine credentials: on engineering laptops the same stealers collect SSH keys, cloud CLI profiles and API keys (see Where identity fits).
  • Hands-on-keyboard intrusion: RATs and RMM tools give operators an interactive foothold for reconnaissance, Kerberoasting and lateral movement, which in the Interlock case led to data theft and encryption.

Detection

Because the user starts the process, detection has to focus on what an ordinary Run dialog or terminal session should never do. Collect the following telemetry and build the analytics in the second table.

Data sourceWhat to collectWhy it matters
Registry: HKCU\Software\Microsoft\Windows\CurrentVersion\Explorer\RunMRUValue writes and current contents per user (EDR registry events or Sysmon Event ID 13)Records commands entered into Run. Microsoft notes it is not written if execution fails, so absence does not prove innocence
Process creation (Security 4688 with command line, Sysmon Event ID 1, EDR)Parent explorer.exe spawning powershell.exe, mshta.exe, cmd.exe, rundll32.exe, wscript.exe, curl.exeThe parent-child pair is the core ClickFix signal
PowerShell script block logging (Event ID 4104)Decoded script content, including Base64 -EncodedCommand payloadsReveals the download cradle after de-obfuscation
Object access (4663) on the WinX folder under %LocalAppData%Access correlated with explorer.exe launching PowerShellUnit 42's indicator for the Win+X variant
DNS and proxy logsRequests to direct IPs, paste sites, URL shorteners, unusual TLDs and newly registered domains straight after a shell launchLinks the paste to the second-stage download
macOS EDRTerminal launching curl or osascript, quarantine attribute removalCovers the AMOS variant

Analytics to build

  • Run dialog to interpreter: explorer.exe as parent of powershell.exe or mshta.exe with a URL, -enc, iex, iwr or irm in the command line. High fidelity in most estates because legitimate users rarely type these into Run.
  • RunMRU content hunt: sweep RunMRU across endpoints for LOLBin names, URLs, Base64 strings or phrases such as "verification" and "captcha" that lure kits append as comments. Microsoft publishes a Defender XDR DeviceRegistryEvents query for this.
  • Shell then network then child: a shell launched interactively that makes an outbound connection within seconds and then spawns certutil, rundll32, msiexec or a binary from a user-writable folder.
  • New persistence after a paste: a Run key or Startup folder item created within minutes of the shell launch. CISA lists a Run key named "Chrome Updater" in Interlock intrusions.
  • Vendor detections: make sure categories such as Behavior:Win32/ClickFix and Trojan:HTML/FakeCaptcha (named by Microsoft) alert rather than silently remediate, because the user may retry on another device.

Metrics that show it is working

  • Percentage of managed Windows devices where the Run dialog is removed or the interpreter launch from Run is blocked (target: all standard user devices).
  • Number of explorer.exe to interpreter alerts per month, and the share triaged within one hour.
  • Share of users who report a fake CAPTCHA simulation rather than following it.

Prevention and hardening

Decision guidance: which controls for which devices

ControlEffect on ClickFixBusiness impactBest fit
Group Policy "Remove Run menu from Start Menu" (User Configuration, Administrative Templates, Start Menu and Taskbar)Removes the Run dialog and the Win+R shortcut, breaking the most common lure instructionsLow for most office staff; IT and developers objectAll standard users; exempt IT through a security group
Application control (App Control for Business or AppLocker) blocking mshta.exe, wscript.exe and PowerShell for standard usersStops the interpreter even if the user pastes into another windowMedium; needs an audit period to find legitimate scriptsMature estates; start with mshta, which few organisations need
PowerShell execution policy AllSigned or RemoteSigned plus script block loggingLogging gives visibility; execution policy alone is not a security boundary because inline commands are not filesLowEveryone, as a baseline
Attack surface reduction rules: block obfuscated scripts, block unproven executables, block JavaScript or VBScript launching downloaded executablesInterrupts common second stagesLow to medium; run in audit mode firstMicrosoft Defender estates
Windows Terminal multiLinePasteWarning and largePasteWarning set to trueAdds a confirmation prompt for multi-line or large pastes (over 5 KiB)NegligibleDevelopers and admins who keep shell access
Protective DNS and web filtering on the device (network protection, SmartScreen, managed browser)Blocks known lure and payload domains wherever the device isLowAll devices, including remote and BYOD browsers

Small organisations without application control tooling get most of the benefit from the Run menu policy, protective DNS and the training rule. Large organisations should treat ClickFix as an application control problem and fold it into the Application Allowlisting Guide work, with script hosts handled as described in the PowerShell Security Guide and the Living off the Land Guide.

Phased rollout

  1. Week 1, visibility: enable script block logging and command-line process auditing, start collecting RunMRU changes, and deploy the explorer.exe to interpreter analytic in alert-only mode.
  2. Weeks 2 to 3, pilot the Run removal: apply the Run menu policy to one business unit. Measure help desk tickets and identify users who genuinely need Run, then move them to an exemption group with stronger monitoring.
  3. Weeks 3 to 6, application control in audit mode: log every launch of mshta.exe, wscript.exe, cscript.exe and PowerShell by non-IT users. Review legitimate uses with application owners.
  4. Weeks 6 to 8, enforce: enforce the Run policy estate-wide and block mshta for standard users, then PowerShell where audit data shows no business need. Keep the previous policy version ready to redeploy as the rollback path if a critical process breaks.
  5. Ongoing, awareness and testing: add fake CAPTCHA pages to phishing simulations and teach the single paste rule with screenshots of real lures. See the Phishing Simulation Programme Guide.

Response when a paste is confirmed

  • Isolate first: network-contain the device before reimaging, preserving memory if the payload is fileless.
  • Assume credentials are gone: reset the user's passwords, revoke all SaaS and identity provider sessions and refresh tokens, and rotate any API keys, SSH keys or cloud credentials stored on the device.
  • Scope by lure: search proxy logs and RunMRU for other users hit by the same lure. Follow the Phishing Incident Response Playbook.

Where identity fits

ClickFix is a delivery mechanism, but its business purpose is almost always credential theft. On a developer or administrator workstation the most valuable loot is non-human: cloud access keys, personal access tokens, kubeconfig files, CI secrets in local environment files and service account passwords. Keep those out of reach of a user-context stealer by using short-lived, federated credentials, hardware-backed keys and a secrets manager rather than plain text files, and make sure your incident runbook includes rotating every machine credential the user could access. The same "paste this to fix it" pattern now targets AI coding agents, which run commands with the developer's credentials; treat agent tool output as untrusted input.

Lessons from incidents

  • Fake prerequisites for agents and users: in the ClawHub Malicious Skills 2026 campaign, malicious agent skills carried a "prerequisites" block telling the user or the agent to run an encoded shell command, which on macOS delivered AMOS. The lesson: paste-and-run lures follow users into every tool that accepts commands, so the paste rule must cover terminals inside IDEs and agents.
  • Fake fixes in tool output: the Sentry MCP Agentjacking 2026 research showed that a crafted error event with a fake "Resolution" section could lead coding agents to run an attacker's package install command with the developer's privileges. It was a controlled test, not a real-world attack, but it is ClickFix with the human replaced by an agent.
  • Ransomware entry through a CAPTCHA: CISA's Interlock advisory shows a paste leading to a RAT, credential theft, RDP lateral movement and exfiltration to cloud storage. The advisory's own mitigations include restricting command-line and scripting use, not only training.

Framework mapping

FrameworkReferenceHow it applies
MITRE ATT&CKT1204.004 Malicious Copy and Paste; T1059.001 PowerShell; T1218.005 Mshta; T1189 Drive-by Compromise; T1566 Phishing; T1105 Ingress Tool Transfer; T1219 Remote Access ToolsThe ClickFix chain from lure to payload
MITRE D3FENDD3-EAL Executable Allowlisting; D3-PSA Process Spawn AnalysisBlocking script hosts and detecting explorer.exe to interpreter chains
NIST SP 800-53 Rev. 5AT-2(3) Social Engineering and Mining; CM-7(5) Authorized Software, Allow-by-exception; SI-3 Malicious Code Protection; SI-4 System Monitoring; AU-12 Audit Record GenerationTraining, allowlisting, malware protection and logging
NIST CSF 2.0PR.AT-01; PR.PS-05; DE.CM-09; DE.AE-02Awareness, preventing unauthorised execution, monitoring and analysis
CIS Controls v8.12.7 Allowlist Authorized Scripts; 8.8 Collect Command-Line Audit Logs; 9.2 Use DNS Filtering Services; 14.2 Train Workforce to Recognise Social Engineering AttacksSafeguards that directly break or reveal the chain
ISO/IEC 27001:2022 Annex AA.6.3 Information security awareness, education and training; A.8.7 Protection against malware; A.8.16 Monitoring activities; A.8.23 Web filteringControl objectives for the same measures

Practitioner checklist

  • Apply "Remove Run menu from Start Menu" to all standard users, with a named exemption group for IT and developers.
  • Block mshta.exe for standard users and plan PowerShell restrictions from application control audit data.
  • Enable PowerShell script block logging (Event ID 4104) and command-line process auditing, and forward both to the SIEM.
  • Deploy the explorer.exe to interpreter detection and a scheduled RunMRU hunt across the estate.
  • Turn on attack surface reduction rules for obfuscated scripts and unproven executables, audit first.
  • Add fake CAPTCHA and fake error lures to awareness training and simulations, with the single paste rule.
  • Make "rotate all credentials and machine secrets on the device" a mandatory step in the infostealer response runbook.

Standards and references

Related NHI Mgmt Group resources: Phishing Guide · Infostealer Threat Guide · PowerShell Security Guide · Browser Security Hardening Guide · Lumma Stealer Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 27 October 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.