Langflow Vulnerability Abused In Attacks Targeting AWS Access Keys

Cybercriminals are no longer just breaking down the front door; they are hijacking the very messaging nervous systems that power modern applications.

On May 5, 2026, the Sysdig Threat Research Team (TRT) uncovered a sophisticated campaign where attackers exploited a critical flaw to turn a high-speed messaging server into a covert command-and-control (C2) hub.

Dubbed “NATS-as-C2,” this novel technique bypasses traditional HTTP channels, allowing operators to orchestrate compromised machines with terrifying efficiency.

The threat actors leveraged CVE-2026-33017, a remote code execution vulnerability in the AI pipeline tool Langflow, to infiltrate networks and steal AWS access keys.

The attack unfolded with precision over roughly 20 hours after the Langflow flaw was added to the CISA Known Exploited Vulnerabilities catalog.

After breaching the unauthenticated instance, the attacker immediately dumped the system’s environment variables, snatching valuable AWS credentials.

Within minutes, they began a massive reconnaissance sweep across the victim’s cloud environment, checking services like S3 buckets, EC2 instances, and Lambda functions.

Langflow Attacks Target AWS

At the heart of this campaign is a custom-built malware project named KeyHunter by the operator. After securing access, the attacker downloaded a specialized Python worker and a heavily armored Go binary to hunt for additional secrets.

Unlike traditional scrapers that target GitHub, KeyHunter is explicitly designed to silently extract credentials from popular online code sandboxes like CodePen, JSFiddle, StackBlitz, and CodeSandbox.

=== KeyHunter Python Worker ===
Worker ID: py-XXXXXX
NATS: nats://45.192.109.25:14222
Capabilities: ['scan_cde', 'scan_web', 'validate_aws', 'validate_ai']
[REGEX] Loaded 12 patterns

Developers frequently leave sensitive API keys in these shared environments while testing snippets, making them an absolute goldmine for credential hunters.

The KeyHunter tool is dangerously sophisticated, built to scrape environments, validate stolen AWS and AI provider keys in real time, and route them back to the attacker’s server.

It uses browser-fingerprint mimicry to disguise its automated traffic as a legitimate web browser, such as Chrome or Safari.

nats: encountered error
nats.errors.Error: nats: permissions violation for publish to "heartbeat.worker"

This allows the malware to slip past advanced bot-detection systems from security vendors such as Cloudflare and Akamai.

If standard web requests fail, the malware deploys a headless browser sidecar to render heavy JavaScript pages, ensuring no stolen credentials are left behind.

According to Sysdig research, the attacker completely ignored basic operational security on the machines hosting their malicious workers.

The automated installation scripts made no effort to hide audit traces, clear system logs, or disguise their systemd persistence mechanisms.

The attackers treat their compromised virtual private servers as cheap, expendable infrastructure.

Rather than carefully hiding a few powerful nodes, they scale their operation by adding numerous disposable machines, utilizing cost-effective ARM instances to keep overhead low while continuously validating harvested keys.

Indicators of Compromise (IOCs)

Type	Indicator
NATS C2	45.192.109.25:14222
Staging HTTP	159.89.205.184:8888

Note: IP addresses and domains are intentionally defanged (e.g., [.]) to prevent accidental resolution or hyperlinking. Re-fang only within controlled threat intelligence platforms such as MISP, VirusTotal, or your SIEM.

Follow us on Google News , LinkedIn and X to Get More Instant Updates. Set Cyberpress as a Preferred Source in Google.

The post Langflow Vulnerability Abused In Attacks Targeting AWS Access Keys appeared first on Cyber Security News.