Remote Support Is Temporary Privileged Access, Not Screen Sharing
A remote support session is a time-bound path to privileged actions on an endpoint. How to bind each session to an accountable identity, keep authority short lived, preserve evidence across the access path and verify that access actually ended.

Guest contribution by Kip Ecklund, founder of Nizlo.
A remote support session gives another person a path to act on an endpoint. The operator may see confidential information, transfer files, run commands, change settings or work through an already authenticated user session. Encryption protects the connection, but it does not answer the questions that matter after an error or security incident: who was authorized, which endpoint was involved, what the operator could do, what actually happened and whether the access path was closed.
That is why organizations should manage remote support as temporary privileged access. The session should be bound to an accountable identity and a defined purpose, limited in duration and capability, recorded across the systems that carry the access, and verified as ended. The same approach helps during routine support, vendor reviews and incident response because it produces evidence that leaders can test instead of relying on a single console event.
Remote access creates authority
The familiar screen-sharing description is incomplete. A screen is the interface, not the control boundary. The real boundary is the authority granted to the operator and the route that authority takes through an account, a service, a relay and the endpoint operating system. A session may allow only viewing, or it may permit keyboard input, command execution, file transfer, clipboard access, service changes and unattended reconnection. Those capabilities create materially different risks even when the user experience looks similar.
CISA, the National Security Agency and the Multi-State Information Sharing and Analysis Center documented this distinction in their joint advisory on malicious use of legitimate remote monitoring and management software. The advisory described actors using portable remote-access executables that did not require a full installation or administrative rights. It also warned that legitimate RMM software can be used for persistence or command and control and may not trigger ordinary antivirus defenses. CISA recommends auditing remote-access tools, reviewing logs for portable execution and using application controls to prevent unauthorized versions from running. [1]
The lesson is not that remote support is inherently unsafe. It is that the connection belongs inside the access-control and incident-response program. An inventory should distinguish authorized products, approved deployment methods and expected support relationships from software that merely happens to provide remote control. Security teams also need a way to identify portable or memory-only execution, because an installed-agent inventory alone can miss it.
Bind access to identity and scope
Every session should begin with an individual operator identity. Shared technician accounts erase attribution at the point where it matters most. If several people can authenticate as the same user, a session record may identify the account while leaving the human operator unknown. That weakens approval, investigation and review, even when the remote-support platform recorded the connection correctly.
Authentication should be followed by an explicit authorization decision. The decision should bind the operator to the customer, endpoint, support purpose and capabilities required for that task. Access to one workstation for a printer problem is not equivalent to access across a customer tenant. Viewing a screen is not equivalent to opening a shell or transferring an executable. High-risk work may justify separating the person who requests access, the person who approves it and the operator who uses it.
NIST Special Publication 800-207 provides a useful model. It says that zero trust does not grant implicit trust because of network location or asset ownership. Authentication and authorization happen before a session to a resource is established, and trust is reevaluated continually while the session is in use. [2] Applied to remote support, that means a successful login is the start of the decision, not the end. Policy should still consider the target endpoint, operator role, customer boundary, device posture, time, requested capability and current risk.
NIST SP 800-53 Revision 5 adds a practical control vocabulary. Account Management, Least Privilege, Remote Access, Event Logging and Audit Record Generation help reviewers ask whether accounts are individually managed, privileges are limited, remote access is authorized, relevant events are defined and systems can create usable records. [3] These controls do not prescribe one remote-support product. They clarify the outcomes an organization should be able to demonstrate.
Keep authority short lived
Temporary access needs visible states. A support workflow should show when access was requested, approved, established, active, suspended and ended. It should also define what ends the authority: operator disconnect, user termination, approval expiry, incident closure, maximum duration or a policy decision. A reusable unattended credential should not remain valid simply because the interactive window closed.
Session-specific authorization is preferable when the workflow supports it. A short-lived token, approval or just-in-time role reduces the value of stolen credentials and narrows the investigation window. Unattended access may still be necessary for managed systems, but it needs an explicit policy basis, stronger authentication, tightly assigned scope, periodic review and a reliable revocation path. User consent should be required when appropriate; when consent is not feasible, the organization should document the policy and contractual basis for access.
Controls also need failure behavior. If the identity provider is unavailable, a relay loses contact or an endpoint is offline, the system should not silently extend privileged access. Emergency access, if permitted, should use a separate accountable process with limited scope, notification and after-action review.
Preserve evidence across the access path
No single log can prove the whole session. Useful evidence comes from at least three layers, and each layer answers different questions.
Control-plane records can identify the authenticated account, role, customer tenant, target device, approval, policy decision, issued session and administrative changes. They can show that a named account received authority at a particular time. They usually cannot prove that the endpoint accepted a command, that a transferred file executed or that a portable access tool was absent from the device.
Relay and session records can show that a connection was established, how long it remained open, which session features were invoked and whether a file transfer or command channel was used. They may also record a disconnect event. These records usually cannot prove that the operating system completed the requested action, that a file was safe or that every reusable credential and endpoint-side token was revoked.
Endpoint evidence can show processes, services, logons, security events, file creation, command execution, privilege changes and security-tool detections. A file hash can distinguish one transferred artifact from another. Endpoint records usually cannot identify the upstream approver or explain why access was granted unless they are correlated with control-plane data.
Correlation therefore matters more than volume. The organization needs a common session identifier, synchronized time, durable operator and endpoint identifiers, and retention long enough for investigation. Material actions such as command execution, file transfer, privilege change and policy override should be reviewable. Recording full session content may create privacy, legal and storage obligations, so the evidence design should be deliberate rather than automatic.
An operator-disconnected event is useful but narrow. It proves that one component observed a disconnect. It does not prove that a service stopped, a token expired, an alternate session closed or the endpoint can no longer be reached. Treating that event as complete closure creates a gap between what the console reports and what the organization can establish.
Revoke, verify and improve
Closure should revoke the session and any temporary authorization, then verify the result from more than one vantage point. The control plane should show the session closed and the approval expired. The relay should show no active channel. The endpoint should show that the expected service, process, token or access path is absent or unusable. Any persistent agent that is supposed to remain should return to its approved state with no orphaned session or unexpected account.
Exceptions deserve their own queue. Shared credentials, stale agents, unrecognized portable tools, orphaned operator accounts, unexplained file transfers and sessions without a matching approval should have an owner and a deadline. Security teams should preserve enough evidence to determine whether an exception was a process failure, a product limitation or malicious activity.
NIST SP 800-61 Revision 3 places incident response within the six functions of the Cybersecurity Framework 2.0. Detect, Respond and Recover handle active incidents, while Govern, Identify and Protect support preparation. The publication also emphasizes continuous improvement rather than waiting until recovery ends to apply lessons learned. [4] Remote-support evidence should therefore improve policy, inventory, detection and access design after every meaningful exception or incident.
Review checklist
- Can every session be tied to an individual operator and a separate, durable endpoint identity?
- Does authorization identify the customer, target, purpose, permitted capabilities and expiration?
- Are unattended access and emergency access governed by stronger, documented controls?
- Can reviewers correlate control-plane, relay and endpoint evidence with synchronized time and a common session identifier?
- Do records distinguish viewing, command execution, file transfer, privilege change and policy override?
- Does the organization inventory installed, portable and memory-only remote-access software?
- When access ends, are temporary approvals, tokens, sessions and endpoint-side paths revoked and independently verified?
- Are exceptions assigned to an owner, investigated and used to improve policy and detection?
A remote-support platform should make these answers easier, but the organization still owns the access model. Treating each session as temporary privileged access connects identity, authorization, evidence and revocation into one accountable workflow. It also gives business and IT leaders a clearer basis for vendor evaluation and incident review than a statement that the screen-sharing window was closed.
For related guidance, see our zero trust implementation strategies, the incident response plan template and the vendor breach hour-zero playbook.
Sources
- CISA, NSA and MS-ISAC, Protecting Against Malicious Use of Remote Monitoring and Management Software, AA23-025A, January 2023
- NIST Special Publication 800-207, Zero Trust Architecture, August 2020
- NIST Special Publication 800-53 Revision 5, Security and Privacy Controls for Information Systems and Organizations
- NIST Special Publication 800-61 Revision 3, Incident Response Recommendations and Considerations for Cybersecurity Risk Management, April 2025
Kip Ecklund is the founder of Nizlo, a Windows-first remote monitoring and management software company. His work focuses on endpoint operations, support workflows and the evidence required to manage remote access responsibly.
Related from the BlueRadius Library
Sourced posts on adjacent topics, ranked by tag overlap.
AI Security
The AI Agent That Writes Its Own Audit Trail Is an Incident Waiting to Happen
Don't let an AI agent's summary close an incident. Reconcile agent traces with access, destination and approval records, and run a forged-record test.
ReadThreat Intelligence
When the Patch Hasn't Landed Yet: A Defender's Playbook for Edge Appliance Compromises
FortiGate and Ivanti intrusions show the same pattern: logs rotate, credentials persist, the patch is not the end. A defender's playbook for the first 24 hours after an appliance z
ReadAI Security
AI Coding Agents Have a Software Supply Chain, Too
Treat MCP servers and coding-agent plugins as dependencies: inventory them, verify provenance, scope credentials and re-review when tools or permissions change.
ReadThreat Intelligence
Detecting Kerberoasting and AS-REP Roasting From Windows Logs: A Practical Detection Playbook
Detect Kerberoasting and AS-REP roasting in Windows events 4768 and 4769: PowerShell hunting queries, alert context and the fixes that stop both attacks.
ReadAI Security
Exploits Now Arrive in Hours. Your Approval Chain Still Takes Weeks.
McKinsey finds critical vulnerabilities are now exploited within hours of disclosure, down from about three weeks. Why decision rights, not tools, set your exposure window, and six
ReadThreat Intelligence
Assume the Vendor Is Compromised: The Hour-Zero Playbook
A vendor breach compromises everyone downstream at once, and you often hear it from a headline first. The playbook to build before the outage, and the four moves that matter in the
ReadRelated services