Even "cat readme.txt" is not safe

This article details a critical vulnerability in iTerm2, demonstrating how a seemingly harmless command like cat readme.txt can be leveraged to execute arbitrary code. The exploit targets iTerm2's SSH integration feature, which, in its effort to provide a richer remote session experience, inadvertently creates a trust boundary issue.

• iTerm2's SSH integration employs a 'conductor' script that communicates with the terminal emulator via terminal escape sequences over a pseudoterminal (PTY).
• The core vulnerability stems from iTerm2's failure to properly validate the source of these protocol messages, accepting forged DCS 2000p and OSC 135 sequences from untrusted terminal output.
• A malicious file, when catted, can impersonate the conductor, tricking iTerm2 into believing it's in an active SSH integration.
• iTerm2 then proceeds with its normal workflow, issuing commands like getshell and pythonversion, to which the malicious output provides forged replies.
• Crucially, the forged DCS 2000p hook includes attacker-controlled sshargs, which iTerm2 later incorporates into a run command payload.
• This allows the attacker to craft a base64-encoded executable path that, when processed by iTerm2, is executed locally by the shell.
• The bug was responsibly disclosed to iTerm2 on March 30th, fixed on March 31st, but at the time of writing, the patch has not yet reached stable releases.
• Notably, the authors mention OpenAI's partnership in this project and their ability to reconstruct the exploit from the commit patch using an LLM.

The research reveals a profound lesson in the complexities of terminal security, where the desire for advanced features can open unexpected doors for exploitation, turning a basic file display command into a critical security risk.