Arista rushes fix for VeloCloud flaw already being used in attacks
A perfect-10 command injection bug in on-premises VeloCloud Orchestrator lets anyone on the network take over the box. Attackers found it first.
Key points
- Arista patched CVE-2026-16812, a maximum-severity 10.0 command injection flaw in on-premises VeloCloud Orchestrator, on Monday.
- The bug lets anyone who can reach the web interface run commands on the server without a username or password.
- Attackers are already exploiting it in the wild, and CISA has ordered U.S. Federal agencies to patch by July 30, 2026.
- Fixed versions are VCO 5.2.3.14, 6.1.3.4, 6.4.2.4 and 7.0.0.1; hosted and dedicated deployments were patched before disclosure.
- Arista published three attacker IP addresses and warns that a breached orchestrator can expose every VeloCloud Edge device it manages.
Arista has shipped an emergency patch for a critical flaw in VeloCloud Orchestrator, the console companies use to run their VeloCloud SD-WAN networks (software that steers office internet traffic between branches). Criminals were already exploiting it before the fix landed.
The bug is tracked as CVE-2026-16812 and carries a severity score of 10.0 out of 10. It's an unauthenticated OS command injection: an attacker who can reach the orchestrator's web page can trick it into running commands on the underlying server. No login required. That pattern should feel familiar, we reported a near-identical pre-auth command injection against Cisco's SD-WAN Manager back on 8 June, where no patch existed at all at disclosure.
The flaw affects on-premises installs of VeloCloud Orchestrator, often shortened to VCO. Arista's hosted and dedicated versions were quietly patched before the advisory went out.
What is VeloCloud Orchestrator and why does this matter?
VCO is the central brain for a company's software-defined wide area network. It configures and pushes policy to all the edge boxes sitting in branch offices. Break the brain and you can reach every limb.
Arista's advisory is direct about the blast radius. A successful attack can hit the confidentiality and availability of the orchestrator and the data it manages, including device inventories, credentials, certificates and cryptographic keys. Attackers can then pivot to the VeloCloud Edge devices themselves.
VCO is exposed by default, and there's no configuration option to hide the vulnerable functionality. If your orchestrator is reachable from the internet, assume someone has already knocked.
Which versions are affected?
| Release train | Fixed in |
|---|---|
| VCO 5.2.x | 5.2.3.14 |
| VCO 6.1.x | 6.1.3.4 |
| VCO 6.4.x | 6.4.2.4 |
| VCO 7.0.x | 7.0.0.1 |
End-of-support versions weren't assessed. Arista tells customers on older trains to contact its technical assistance team about upgrades. If you're running unsupported VCO, you don't know whether you're safe, and you probably aren't.
CISA added the flaw to its Known Exploited Vulnerabilities catalog and gave federal civilian agencies until July 30, 2026 to patch or take the software offline, under Binding Operational Directive 22-01.
How would defenders spot an attack?
Arista published three IP addresses seen exploiting the bug: 8.19.75.217, 206.72.242.124 and 206.72.242.162. Block them, then check logs for past connections. The company is careful to say this list isn't definitive, so clean logs aren't proof of safety.
Administrators should also look for unusual web requests with encoded characters or references to internal services, unexpected outbound HTTPS traffic from the VCO host, and unauthorised config changes or database exports appearing where they shouldn't.
One uncomfortable point on the identity side. This is a pre-auth flaw, so multi-factor authentication on the admin console wouldn't have stopped it. An attacker never touches the login page. Network-layer restriction matters more than usual here: keeping VCO off the public internet and limiting access to a management network is the meaningful control.
What if a VCO instance was already breached?
Patching alone isn't enough for a system that was already hit. Arista's guidance: preserve logs and filesystem timestamps first, then rotate every credential the orchestrator held, review administrator activity, validate every managed edge device, and consider rebuilding the VCO instance from scratch. Certificates and keys stored on the box should be treated as burned.
