Skip to content
NewCybercrime Report 2026 · Eight Chapters · A Post Is Not An Incident → Explore The ChaptersCybercrime Report 2026 → Explore

Odin Research / What is / Tunnels and remote access

What Is noVNC? The Browser VNC Client and Its Phishing Abuse

A harmless remote-desktop viewer that runs in any browser, and a building block for phishing pages that hand victims a real browser on the attacker's server.

By
Odin by Brandefense
Updated
21 Nov 2026
Reading time
6 min read
Type
Remote access

In short

noVNC is an open-source VNC client written in HTML and JavaScript that runs in a web browser. It connects to a remote desktop through a WebSocket proxy such as websockify. Cloud and virtualization consoles use it widely. Phishers can abuse it to show victims a real browser running on an attacker's server.

Key takeaways

What it is
An HTML VNC client and JavaScript library, licensed mainly under MPL 2.0, that lets a browser view and control a remote desktop.
How it works
The browser talks WebSocket to a proxy such as websockify, which forwards the traffic to a normal VNC server over TCP.
Why it matters
A published phishing technique uses noVNC to let victims log in through the attacker's own browser, so MFA is completed on the attacker's machine.
What Odin Research saw
One server in Monograph 06 paired a confirmed mitmproxy listener with a complete noVNC client, 117 files in all. No live session or victim was seen.
What to do first
Use phishing-resistant, origin-bound passkeys, and alert on noVNC or websockify started on cloud servers outside approved VDI.

What is noVNC?

noVNC is an open-source VNC client built from HTML and JavaScript, so it runs in any modern web browser without a plug-in. It lets a user view and control a remote desktop from a web page. Virtualization and cloud consoles rely on it, and phishers have shown how to abuse it.

The project lives on GitHub as novnc/noVNC. Its README describes it as both a JavaScript library and an application built on that library, licensed mainly under the Mozilla Public License 2.0. The README lists projects that use it, including OpenStack, OpenNebula, LibVNCServer and ThinLinc.

VNC (Virtual Network Computing) is an old, simple protocol for sharing a screen and sending keyboard and mouse input. noVNC brings it into the browser. On its own, noVNC is not malicious. It is a normal part of many remote desktop and VDI setups.

How noVNC works

Browsers cannot open raw TCP sockets, and most VNC servers do not speak WebSocket. noVNC bridges that gap with a proxy:

  1. The browser loads the noVNC web page and its JavaScript, such as vnc.html, core/rfb.js and app/ui.js.
  2. The WebSocket proxy, usually the sister project websockify, accepts the browser's WebSocket connection and forwards it to a VNC server over TCP. The bundled novnc_proxy script downloads and starts websockify, which also serves the web files.
  3. The VNC server shares a real desktop or a single application, such as a browser.

The README lists the oldest browser versions it is aware of, such as Chrome 89, Firefox 89 and Safari 15. Its examples run the proxy on port 6081, but any port can be used.

How attackers abuse noVNC for phishing

In February 2022, the researcher mr.d0x published a technique to steal credentials and bypass two-factor authentication with noVNC. The idea is simple:

  1. The attacker runs a browser in kiosk mode on a server they control and opens a real login page in it.
  2. noVNC shows that browser to the victim through a link, hiding the VNC interface so it looks like a normal web page.
  3. The victim types the password and the MFA code into what is really the attacker's browser.
  4. The attacker keeps the logged-in session and its cookies after the victim leaves.

The author said the technique was developed because some sites had started to block reverse-proxy phishing tools such as Evilginx. It fits the wider adversary-in-the-middle family and the ATT&CK techniques Adversary-in-the-Middle (T1557) and Steal Web Session Cookie (T1539).

A noVNC lure defeats one-time codes and push approvals, because the real site sees a real browser completing a real login. It is weaker against origin-bound passkeys, which check the site the user's own browser is talking to.

What Odin Research observed

Monograph 06 reviewed 2,750 phishing, proxy and tunnel listeners and three exposed workbenches. One of them held noVNC.

  • A proxy beside a noVNC client. Host B, a VPS at a small hosting provider in India, ran a confirmed mitmproxy TLS listener. On 13 November 2026 the same address served an open directory of 117 files (917,101 bytes): a complete stock noVNC 1.x client with core/rfb.js, app/ui.js, core/websock.js, vnc.html and the novnc_proxy helper.
  • One part of a possible replay setup. Replaying a stolen cookie from a browser on the proxy server would keep the victim's session on one address. That would also need a running browser, a VNC server, a WebSocket bridge and a cookie transfer. None was observed, and no login lure or victim session was found.
  • Watchlist, not block list. Host B stays on a watchlist with its address withheld. Odin also warns that stock noVNC file hashes must never be blocked alone, because they appear in legitimate Proxmox VE, OpenStack and VDI consoles.

The same monograph covers Ligolo-ng tunnels and a keystroke-capture kit beside Evilginx2, which together show how session theft and network access fit together.

How to detect noVNC abuse

  • noVNC pages on unexpected hosts. Scan for HTTP 200 responses that serve vnc.html or core/rfb.js outside your approved VDI and hypervisor consoles.
  • Raw VNC greetings. A VNC server answers with an RFB 003.00x banner. One on an internet-facing cloud host deserves a look.
  • Processes on cloud VMs. Alert when noVNC or websockify starts on cloud or VPS hosts that have no remote desktop role.
  • Sign-in context. In your identity logs, look for successful logins that come from a hosting provider address right after a user clicked an unusual link.

Odin Research publishes rules for these signals. Monograph 06 carries Suricata signatures for noVNC client pages and raw VNC server greetings, and a Sigma rule for noVNC on cloud VMs. Exclude sanctioned VDI first. The rules are on the report page and in the Defense Kit.

How to defend against noVNC phishing

  1. Move to passkeys. Origin-bound WebAuthn passkeys, combined with device compliance checks, are the strongest control against proxy and remote-browser phishing. See passkeys.
  2. Bind sessions to devices. Conditional access that requires a managed device stops a session that lives on an attacker's server from reaching sensitive apps.
  3. Hunt for replayed sessions. Treat session cookies as credentials. See session hijacking and Phishing 101.
  4. Keep legitimate consoles private. Put your own noVNC consoles behind VPN or SSO, so a public noVNC page in your space is always suspicious.

noVNC compared with other phishing building blocks

ToolRole in phishingDefeats OTP and push MFA?Legitimate use
noVNCShows the victim a real browser on the attacker's serverYesVDI, hypervisor and cloud consoles
EvilginxReverse proxy that relays the real login pageYesRed-team testing
mitmproxyIntercepting proxy that can read TLS trafficOnly with a trusted root CA or proxy setupDebugging and testing
GoPhishSends lures and tracks clicksNo, on its ownPhishing awareness tests

Frequently asked questions about noVNC

What is noVNC?

noVNC is an open-source VNC client written in HTML and JavaScript that runs in a web browser. It connects to a VNC server through a WebSocket proxy such as websockify.

Is noVNC malware?

No. noVNC is a legitimate open-source project used by OpenStack, OpenNebula, ThinLinc and many VDI consoles. Attackers can abuse it, but the software itself is not malicious.

Is noVNC safe to use?

Yes, when you run it behind authentication and on approved hosts. Do not expose a noVNC console to the internet without a VPN, SSO or strong access controls.

What is websockify?

websockify is the sister project of noVNC. It accepts WebSocket connections from the browser and forwards them to a normal VNC server over TCP, and can also serve the noVNC web files.

How does noVNC phishing bypass MFA?

The victim logs in, including the MFA step, inside a browser that actually runs on the attacker's server. The attacker then keeps that authenticated session. mr.d0x published the technique in February 2022.

How do I detect noVNC phishing?

Look for noVNC pages such as vnc.html or core/rfb.js on unexpected hosts, raw VNC greetings on cloud servers, and websockify processes on VPS hosts. Odin Research publishes Suricata and Sigma rules for these in its Defense Kit.

Do passkeys stop noVNC phishing?

Origin-bound passkeys are the strongest defense, because the passkey checks the site the user's own browser is talking to. Pair them with device-bound conditional access.

noVNC vs Evilginx: what is the difference?

Evilginx relays the real login page through a reverse proxy. A noVNC lure shows the victim an entire remote browser instead. Both aim to capture an authenticated session.

Sources

  1. noVNC project, README, GitHub. raw.githubusercontent.com/novnc/noVNC/master/README.md (checked 21 November 2026)
  2. mr.d0x, "Steal Credentials & Bypass 2FA Using noVNC", 19 February 2022. mrd0x.com/bypass-2fa-using-novnc/ (checked 21 November 2026)
  3. MITRE ATT&CK, T1557 Adversary-in-the-Middle. attack.mitre.org/techniques/T1557/ (checked 21 November 2026)
  4. MITRE ATT&CK, T1539 Steal Web Session Cookie. attack.mitre.org/techniques/T1539/ (checked 21 November 2026)

Related tools and terms

All Odin Research explanations · Brandefense glossary

Odin Research

Get the rules, not just the definition

Every Odin Research report ships with verified indicators and detection rules you can load today.

Open the Defense Kit