How-to
Making Remote HMI Access Read-Only Over RDP
Short answer
RDP forwards keyboard and pointer input, so no setting inside the remote session makes a PanelView read-only. Replace the launch chain: publish a read-only FactoryTalk ViewPoint client, or enforce view-only mode in the VNC server, then verify from the operator account that every visible control leaves the controller untouched.
On this page
RDP cannot be made read-only. The protocol carries keyboard and pointer input to the logged-on session, so a DCS button that opens the runtime over RDP hands the remote user every control a local operator has. The fix is a different path, not a setting: publish the displays through a read-only FactoryTalk ViewPoint client in a browser, or serve the screen from a VNC server with view-only mode enforced on the server, and keep command authority at the field HMI. Until one of those passes a command-rejection test, disable the button — an interactive session meant for troubleshooting is a live control seat.
Why the RDP layer cannot be hardened #
Walk the chain as it runs today: the DCS button executes a script that logs on to the server with stored credentials, Task Scheduler launches a .bat file that opens the .mer runtime, and the operator then signs in to the HMI to navigate. Each layer answers a different question, and none of them answers who may command the equipment.
- The Windows account governs files and applications in the session, not the pushbuttons inside a running .mer.
- The scheduled task changes how the runtime starts, not what input it accepts.
- The HMI login carries the HMI's own command rights; if that identity may operate equipment, the remote user inherits those rights after sign-in.
RDP sits underneath all three and transports the desktop, input included. The same is true in the other direction: an RDP session from a Comfort Panel to a cabinet PC is interactive by design, which is what makes it useful for operation and wrong for observation. Decision: stop stacking Windows restrictions on this chain and replace it.
Separate observation from command #
The architectural fault is that one path carries both visualization and control. Build two:
- Control stays at the field HMI, where the operator has line of sight to the machine — that is the safety boundary, and it is why a remote start command is never acceptable from a troubleshooting view.
- Observation is published through a session that rejects input or contains only read-only content, reached on its own credential.
Give the observation path its own identity. A credential embedded in the DCS script makes every launch look like the same user, which destroys accountability; the IEC 62443 series expects access to be granted by role, limited to what that role needs, and attributable to an individual, and one stored login in a button script meets none of the three. As defense in depth, create an observer login in the runtime's user security that can navigate but holds no command authorizations, so even a fallback interactive path carries no command rights. The split is not tied to one platform: weighing remote HMI options for a WinCC Unified runtime raises the same observation-versus-control question.
Replace the chain step by step #
- Map the launch chain. Record the DCS script, the stored Windows account, the scheduled task, the .bat file, the .mer target, and the HMI login the session reaches. Mark every point where a credential or a user input enters.
- Contain the exposure now. Disable the button or its stored credential until the replacement passes testing; an interactive route left live while you build the replacement keeps the control seat open.
- Publish a read-only browser session. Deploy FactoryTalk ViewPoint on the PanelView Plus terminal at the machine or on the server, and license the read-only client option. The standard ViewPoint web client is read-and-write, so read-only behavior comes from that option, not from a default install. Publish only the displays troubleshooting needs: status, alarms, trends.
- Enforce view-only VNC where that is the terminal's capability. RealVNC Server documents a parameter that governs how legacy viewers connect: value 1 admits them in view-only mode, and value 2 lets them control whenever no other user has control. Set 1. Value 2 fails the requirement exactly during a solo remote session, which is when this button gets used.
- Create the observer identity. A dedicated Windows account for the path, and where navigation requires an HMI login, an HMI user in the no-command group described above. Never reuse the field operator or maintenance identity.
- Repoint the DCS button. It opens the read-only web page or the approved view-only client directly. Remove the automatic RDP logon, the scheduled launch, and the command-capable HMI login from the operator path.
- Retire the old route. Delete the stored credential and remove the scheduled task entry so RDP cannot come back as a fallback.
Prove it with a command-rejection test #
Test from the production DCS workstation with the exact account the button uses; an administrator test proves nothing about the operator path. Open every published display, exercise every command object by pointer and by keyboard, and watch the corresponding controller tags from an independent trusted station.
Worked example: the published set has 4 displays. Overview carries 5 pushbuttons, the pump screen 3 selectors and 2 numeric entry fields, the alarm banner 1 acknowledge button, and the trend display none. Command objects: 5 + 3 + 2 + 1 = 11. Each exercised once by pointer and once by keyboard: 11 x 2 = 22 input attempts. Close and reopen the session, then press each object once more: 22 + 11 = 33 attempts. The independent station watches the 11 command tags for the whole test and records 0 state changes. A pass is 33 of 33 attempts rejected with 0 tag changes; one change fails the path and the button stays disabled. Record the account, client, displays, attempts, and tag evidence as the baseline.
The decision for each case #
- ViewPoint available on your platform: the read-only client in a browser is the primary answer; it replaces the entire RDP chain, logon script and scheduled task included.
- No ViewPoint, but the terminal runs a VNC server: enforced view-only mode is acceptable once the test above shows input is rejected at the server.
- Neither is deployable: remove the button. Losing the troubleshooting view is the smaller loss next to an uncontrolled remote control seat, and it is restored the day a read-only path exists.
- In every case: commands remain at the field HMI, and the observer identity holds no command rights even after the change.
Questions asked
Does turning on view-only in the VNC viewer give the same protection?
No. View-only set in the viewer is a client-side preference: the same server reached with another client, or the same viewer with the box unticked, regains control. Enforcement has to come from the server parameter that admits connections in view-only mode, because that setting applies to every client that connects, including one an operator installs on their own PC.
If the HMI login is still needed for navigation, can its rights make the session safe?
Only as defense in depth, never as the whole answer. Put the remote login in a group whose members can navigate but hold no command authorizations, and the HMI refuses their commands even though input still arrives over the session. The guarantee depends on every pushbutton, selector, and entry field actually carrying its authorization check, so confirm it with the command-rejection test rather than trusting the group name.
How often should the command-rejection test be repeated?
Repeat it after any change to the HMI application, the user database, the server, or the remote-access setup. A new display can introduce a command object nobody tested, and a server rebuild or firmware update can reset the VNC parameter back to its default. Keep the recorded baseline of account, client, displays, attempts, and tag results so the next run is a comparison, not a first attempt.
Is hiding the control screens an acceptable substitute?
No. Hidden navigation is not an authorization boundary: another navigation path, a pop-up, or a reused display can expose the active objects again, and one route back to the screen restores full operation. A boundary is a session that rejects input at the server or an HMI identity with no command rights; anything less is layout, not security.