Marimo RCE (CVE-2026-39987) exploited in under 10 hours after vulnerability revealed.
Apr 10, 2026 // 13:42 - Norina Velotta


A serious security flaw in Marimo, a free and open Python tool for data work, was targeted just 10 hours after it was revealed publicly, according to Sysdig’s research.

The issue, labeled CVE-2026-39987 (with a CVSS score of 9.3), is a way to run code on a system without proving who you are. It affects all Marimo versions up to and including 0.20.4, but it has been fixed in version 0.23.0.

“The /terminal/ws terminal point doesn’t check login details. This lets someone who isn’t authorized get a full command-line shell and run any system commands,” Marimo’s developers explained in an announcement earlier this week.

“While other areas (like /ws) check user credentials with validate_auth(), the /terminal/ws area only looks at the mode and whether the system supports it before accepting a connection, completely skipping the user check.”

Basically, hackers can get a fully interactive command line on any exposed Marimo system through one WebSocket connection, without needing usernames or passwords.

Sysdig discovered that adversaries attempted to take advantage of the flaw within 9 hours and 41 minutes after its public announcement. The criminals immediately started stealing login credentials, even though no working code (PoC) for the breach was publicly available at the time.

The unidentified threat agent is reported to have connected to the /terminal/ws communication endpoint on a test system and began exploring the system and searching files. In just minutes, they tried to steal data from the .env file, look for SSH keys, and examine different files.

The intruder returned an hour later to the test system to access the contents of the .env file and see if other threat actors were active during that period. No other malicious software like crypto miners or backdoors, were installed.

“The perpetrator created a working exploit directly from the recommendation’s explanation, connected to the open terminal access point, and began personally exploring the affected environment,” the cloud security business said. “The attacker connected four times over 90 minutes, with rests between sessions. This matches a human operator checking through a list of targets and coming back to check results.”

The rapid speed that newly discovered weaknesses are being used indicates that malicious actors are closely monitoring the notifications of loopholes/weak points and quickly using them in attack methods during the period of time in between discovery and application of a fix by the user. Because of this, it shortens the time frame defenders have to react after a vulnerability is made known.

“The assumption that only widely used platforms are targeted is untrue. Any application connected to the internet with a serious warning is a target, despite their popularity.”

#(cve-2026-39987)  #after  #exploited  #hours  #marimo  #news  #rce  #revealed.  #under  #vulnerability   —   News