Description

A critical security flaw in Langflow, a popular Python-based framework for AI application development, has been exploited in a widespread cyberattack campaign to deliver the Flodrix botnet. Assigned the identifier CVE-2025-3248 and carrying a CVSS score of 9.8, this critical vulnerability affects all Langflow releases earlier than version 1.3.0. The flaw resides in the `/api/v1/validate/code` endpoint, which fails to properly validate input, allowing unauthenticated attackers to execute arbitrary Python code via crafted POST requests. With over 1,600 Langflow instances exposed globally, attackers rapidly weaponized the vulnerability using public reconnaissance tools like Shodan and FOFA to find vulnerable systems. Exploits freely available on GitHub enabled threat actors to gain shell access and perform system commands, such as `whoami` and `ip addr show`, gathering intelligence before communicating back to their command-and-control (C2) infrastructure. After establishing control, a malicious script—often camouflaged as a benign utility like “docker”—delivers the Flodrix botnet over TCP or via the Tor network. Flodrix, a descendant of the LeetHozer malware family, is engineered primarily for distributed denial-of-service (DDoS) attacks, using modes like `tcpraw` and `udpplain`. It employs advanced stealth techniques, including XOR-based obfuscation, forensic cleanup, self-deletion, and fake process names, to avoid detection and ensure persistence. To prevent reinfection, Flodrix checks for specific hidden files and deletes its traces post-installation. Its capabilities extend beyond DDoS, posing risks of data theft and service outages for organizations relying on Langflow. Security experts urge immediate updates to Langflow version 1.3.0 or newer, which includes stricter endpoint authentication. Restricting access to Langflow interfaces and monitoring for suspicious activity—such as unusual traffic or hidden files—is also critical. This incident underscores the urgent need for proactive vulnerability management and hardened security practices, particularly when using popular open-source platforms.