Hackers are using malvertising and AI platforms to spread ClickFix attacks, tricking users into running malicious commands and exposing credentials.
TWCERT/CC指出 that recent ClickFix social engineering attacks continue to intensify, with attackers reaching users through malvertising, phishing emails, fake verification pages, or compromised websites, and then using scenarios such as system errors, software installation, or identity verification to lure victims into copying and executing malicious commands on their own.[1] The key point of this method is to induce users to actively execute malicious code.[1]
Trend Micro TrendAI™ Research observed that attackers have combined ClickFix with Google Ads and legitimate AI platforms, impersonating popular tools such as Claude AI, ChatGPT Codex, Cursor IDE, JetBrains, and Perplexity, and using paid search ads to direct users to fake websites.[1] Some campaigns even abuse Claude Shared Chat by creating public sharing pages on the official platform and disguising them as technical support or development teams, further reducing victims’ caution.[1]
Because the attack content may appear on legitimate platforms or domains that users are familiar with, users are easily less alert due to trust in the platform or brand.[1]
The ClickFix workflow typically first attracts users with fake CAPTCHA prompts, system error messages, or technical support prompts, then asks them to open command-line tools such as PowerShell or Terminal, copy the specified content, paste it, and execute it.[1] MITRE ATT&CK classifies this behavior as T1204.004 "User Execution: Malicious Copy and Paste".[1]
Attackers often use Base64 encoding or command obfuscation to reduce the chance of detection, then induce users to use system tools such as PowerShell, curl, or bash to download subsequent malware.[1] This technique has been widely used to distribute information-stealing malware and other malicious software, and has also appeared in ransomware attack chains.[1]
As generative AI and AI development tools become rapidly more widespread, related services are also increasingly becoming bait or carriers for attackers to build social engineering scenarios.[1]
TWCERT/CC monitoring data shows that about 67% of traffic from campaigns using AI development tools as bait is concentrated in the Asia-Pacific region, with Taiwan accounting for 30.5%, the highest among all countries.[1]
Once a user executes a malicious command, the device may further download information-stealing malware, remote access tools, or other malicious software, and steal browser account passwords, cookies, sessions, and other authentication information.[1] If an enterprise user’s credentials are stolen, it may lead to account impersonation, unauthorized access, and follow-on intrusion risks.[1]
First, remain cautious of web content that asks users to copy and execute PowerShell, Terminal, or other command-line instructions. Even if the content appears on a legitimate or familiar platform, users should still verify the source, content, and actual purpose of the command, and avoid judging safety based only on the URL or brand.[1]
Second, software and development tools should be obtained preferentially through official websites or trusted channels, and users should avoid directly clicking paid search ads or downloading and running installers from unknown websites.[1]
Third, government agencies and enterprises should strengthen awareness campaigns about ClickFix social engineering attacks so that personnel understand that attackers may use popular AI tools, search ads, and legitimate platforms to create trusted scenarios and induce users to execute malicious commands.[1]
Fourth, continuously monitor endpoint command and script execution activity, and watch for Base64 decoding, obfuscated commands, remote content downloads, and unusual use of tools such as PowerShell, Terminal, curl, and bash.[1]
Fifth, if it is discovered that a user has executed an unknown command, it is recommended to immediately isolate the relevant device and conduct an incident investigation, while also checking whether accounts, login sessions, and important authentication information show abnormal use; if necessary, passwords should be reset and existing login sessions revoked.[1]