Node-RED: CVSS 10 flaw exploited to drop Linux bot and webshells
CVE-2026-58115: Node-RED instances without authentication exploited with a single request to install a Linux bot and webshells. Details, IoCs, actions.

Since 30 October 2026, CVE-2026-58115 has been actively exploited. It is the missing authentication on the Node-RED HTTP interface that Siemens disclosed on 11 August with a CVSS score of 10.0. Vulnerability intelligence company VulnCheck spotted it: its decoy systems (canaries) received a complete exploit that installs a Linux bot, and over the following days the same source also left persistent webshells. The Register, SecurityWeek, Help Net Security and CSO Online reported the story and agree on the main facts.
The vulnerability
Node-RED is a flow-based programming tool built on Node.js, with a browser-based editor. It is widely used in manufacturing, industrial automation, healthcare and home automation, and it also ships inside commercial ICS/OT products. Siemens assigned the CVE for its SIMATIC IoT2050 Advanced gateways running Industrial OS with Node-RED installed: on these devices the Node-RED HTTP interface does not require authentication and gives access to programming nodes that run system commands. According to the Siemens advisory, an unauthenticated remote attacker can create malicious flows and execute arbitrary code on the underlying system with maximum privileges. Versions before V4.3.4.1, which contains the fix, are vulnerable.
The problem is not limited to Siemens devices. As CSO Online points out, Node-RED ships with editor authentication disabled: until the administrator configures adminAuth, anyone who can reach the editor can deploy flows. According to VulnCheck, attackers are looking for exactly these instances without adminAuth, wherever they are installed, and exploit them automatically with a single HTTP request.
The observed attack chain
The request captured on 30 October was a complete exploit: a single POST /flows to the admin API carrying both the flow and the dropper, with the command hidden behind three layers of encoding. The flow uses two standard Node-RED nodes:
- an inject node, which fires a message on a schedule or once at startup;
- an exec node, which runs a shell command.
Wired together, they run the command as soon as the flow is deployed, with no further interaction. The command downloaded a file named bot-linux-amd64 from a separate server, saved it under a hidden name in a memory-backed location and ran it in the background, without checking whether it succeeded. VulnCheck did not retrieve the binary, so it cannot say whether it is a miner, a DDoS bot, a reverse shell or something else; the name suggests builds for several architectures.
By 1 November the same source had hit the vulnerability 16 times, with different payloads: a shell script from the same download server, a file from a second server saved under a name mimicking a kernel worker thread and, in several cases, webshells that let the attackers come back and run commands at will. The bot lives in memory and disappears on reboot unless persistence is set up some other way; the webshells stay.
In MITRE ATT&CK terms, the described behaviour maps to T1190 (exploit public-facing application), T1059.004 (Unix shell), T1027 (obfuscation), T1105 (ingress tool transfer), T1036 (masquerading) and T1505.003 (web shell).
Public indicators and context
- 4.201.105[.]116: attack source, in Microsoft’s network (AS8075), geolocated to Brazil;
- 85.215.73[.]250: second server from which one of the payloads fetched the malware;
- file name: bot-linux-amd64.
The VulnCheck post lists the full indicators, including both payload servers. Across VulnCheck’s canary network the same address logged 125 attempts against eight different CVEs, old and new, in Apache Solr (CVE-2019-17558), Camunda, Gitea, MLflow, Traefik and Joomla, as well as Node-RED. The researchers describe it as opportunistic mass scanning, not a targeted attack. An address in a large cloud provider’s range, VulnCheck notes, is usually a rented or compromised virtual machine and says little about who is behind it; payloads coming from servers other than the attacking one point to a “scan and infect” setup.
VulnCheck added the flaw to its own KEV catalogue; when the articles came out (2–3 November) it was not yet in CISA’s KEV catalogue.
Impact for businesses
The observed attacks do not seem interested in industrial processes: they look for any system able to run commands. The analysts quoted by CSO Online stress, however, that this does not lower the risk. Node-RED often acts as middleware between controllers (Modbus, OPC UA, MQTT, S7) and databases, brokers and cloud services, and stores their credentials: a compromised instance becomes a credential vault and a bridge between IT and OT. Even generic malware that consumes resources or disrupts communications can have operational consequences. On top of that, OT firmware updates follow planned maintenance windows, so many gateways will remain vulnerable for a long time.
For European and Italian organisations subject to NIS2, especially in manufacturing, healthcare and critical infrastructure, an exposed Node-RED is an inventory problem even before it is a patching problem: it is often installed by technicians outside IT and does not appear in asset lists.
What to do now
- Inventory: find every Node-RED instance, including those inside gateways and OT products (the editor’s default port is 1880).
- Exposure: check today whether they are reachable from the internet. The editor must not be: access only from management networks or via VPN.
- Authentication: enable
adminAuthon every instance and check that the admin API (/flows) requires credentials. - Updates: on SIMATIC IoT2050 Advanced install V4.3.4.1 or later; for other products that embed Node-RED, follow the vendor’s advisories.
- Detection: search logs for
POST /flowsrequests from external addresses and for connections to the published indicators; review deployed flows for unexpected exec nodes or HTTP endpoints, as VulnCheck recommends; check for processes running from memory-backed locations or with names mimicking kernel threads. - If you find traces: patching is not enough, because webshells survive reboots and updates. Remove the malicious flows, rebuild the system if possible and rotate every credential stored in Node-RED.
How ARXDOME helps
ARX Expose shows which of your company’s services are visible from the internet and the known vulnerabilities on public ones. ARX Scout verifies exposed ports and services first-hand, with your authorisation, and alerts you when a new one opens, such as a Node-RED editor left reachable. ARX Radar shows who is attacking your public addresses right now and, with ARX Horizon, you can search the published indicators of compromise from the portal.
Last updated:


