Skip to content
nulltapE-reader edition

← All e-reader articles

4 min read

Elastic Names ScreenConnect Most Often in 23 RMM Abuse Findings

Elastic's quarterly report describes attackers using signed remote-management agents for interactive access; defenders should verify who deployed each agent.

Read on the standard site

An office computer shows a nested remote desktop window with two pointers linked across the screen.

ScreenConnect appeared most often among the remote monitoring and management (RMM) products in 23 confirmed abuse findings reported by Elastic Security Labs for April–September 2026. For teams that allow remote support on employee machines, an approved agent can give an intruder interactive access under the cover of familiar software.

Use the 23 as a count of findings when scoping your own investigation, rather than as a count of ScreenConnect deployments or affected organizations. Elastic draws the quarterly report from confirmed customer intrusions and first-party research. Its summary groups RMM agents with AI coding assistants and Microsoft first-party clients as legitimate products attackers used for persistent, interactive access. The common problem is an authorized program doing work for the wrong person.

Signed remote tools carry unauthorized sessions

An RMM agent is supposed to let support staff connect to an endpoint and fix it. Once the agent is installed and enrolled, an operator can control the machine remotely. If an attacker gets an agent onto a machine or takes over an existing support path, its signature and product name still look like normal IT software. Verify who approved the installation and who opened the remote session.

Elastic writes, “Out of 23 confirmed RMM abuse findings this quarter, ScreenConnect was the product named most often.” Start by matching every local RMM deployment to the approved support-tool inventory. Investigate an unexpected deployment or session in its own context; your support team may use the same product legitimately.

Elastic also describes a separate healthcare finding in which a paste-and-run lure progressed to Microsoft Software Installer (MSI) delivery and dynamic link library (DLL) sideloading in roughly 16 seconds. Investigate the healthcare MSI sequence separately from the ScreenConnect finding. Look for installer activity and subsequent module loads if your endpoint telemetry captures them; investigate the initiating user, process ancestry and approval record together.

Compare agent inventory with support records

Start with the endpoints where an RMM product is present. Export the installed applications, services and software deployment history, and compare each agent instance with your approved support-tool list, deployment tickets and management-console enrollment records. Ask the service owner whether the endpoint and controlling account belong to an authorized session. A trusted signature establishes the software publisher. Match the controlling account to your team’s support record to establish authorization for this session.

For an unexplained agent or connection, preserve endpoint process starts, installer activity, service creation and RMM console session history where available. Correlate their timestamps with user sign-ins and support tickets. A newly enrolled workstation, a support session outside an approved change, or an unfamiliar controlling account deserves containment and credential review under your incident process. Restrict agent deployment and console access to approved administrators, and remove unauthorized installations through your managed software process.

Use the operator and approval record alongside the product name. Existing agents and legitimate remote help generate the same broad process names. The approval record and operator identity separate routine support from an unapproved session; missing session telemetry means you should preserve the endpoint evidence and ask the tool owner for console-side history before clearing the event.

Elastic’s report also describes attackers using trusted applications and identity flows to reach cloud and software-as-a-service tenants. That is a different route from an RMM session, but it asks the same operational question: who granted a familiar application access, and what can it reach? Elastic describes Adversary-in-the-Middle phishing and legacy authentication in this identity discussion. It says some cases averaged eight minutes from infostealer infection to an attacker sign-in with the same credentials; that timing belongs to those cases, not to the 23 RMM findings.

For Microsoft Entra tenants, Microsoft’s user-consent guidance provides a specific setting to review: Identity > Applications > Enterprise apps > Consent and permissions > User consent settings. Microsoft recommends limiting user consent to apps from verified publishers. Its built-in microsoft-user-default-low policy limits consent to verified publishers and tenant-registered apps for permissions classified as low impact; the microsoft-user-default-legacy policy permits users to grant any permission that does not require admin consent. Review the current policy with your identity team and restrict it where the business allows. Confirm the saved setting and test a permitted low-impact application request and a request that requires administrator consent with designated test accounts.

Changing user consent settings affects only future consent operations. Existing application grants remain in force; review enterprise-app permissions and revoke grants you find unauthorized.

Verify the operator behind the trusted software

Match each deployed remote-support agent to an approved endpoint, controlling account and support record; investigate unknown enrollments using endpoint installer and service events plus console session history. In Entra, record the current user-consent policy, choose a narrower approved policy where appropriate, and test its effect with designated accounts. The first check tests who controls a trusted endpoint tool; the second tests who can grant a trusted application tenant access. Keep the two sets of findings separate when you escalate: Elastic’s RMM count and its identity timing describe different observations.

Primary sources