• Nov 13

MFA Didn't Fail. It Was Never Asked: What the ScreenConnect RMM+ Attack Teaches About Integration Keys

The ScreenConnect RMM+ attack went around MFA through a blank extension key. Why integration keys are credentials, and how MSPs and clients lock them down.

Picture the MSP that looks after your office computers. Every technician signs in to their remote-access console with a strong password and an authenticator app. Roles are set up so the junior tech can't touch the payroll server. On paper, it's a well-run shop.

On the afternoon of November 11, that's roughly what several MSPs were looking at. Commands landed on client machines across several ScreenConnect instances within minutes, and in the logs Huntress reviewed there was no corresponding login activity at all. No technician signed in, so no MFA prompt ever fired.

According to Huntress, which investigated the activity, the commands came through a third-party ScreenConnect extension called RMM+. The extension lets another tool drive ScreenConnect sessions, and it checks one thing before it obeys: a shared access key. In two versions, if nobody had ever set that key, a blank one was accepted.

So MFA didn't fail. Nobody ever asked for it. The attacker didn't come through the front door where MFA stands guard. They came through a side door that only checked a key, and the key slot was empty.

This post explains what happened in plain English, why "integration keys" are credentials that need the same care as admin passwords, and what MSPs and their small-business clients should do this week.

Everything here was checked against the ConnectWise Trust Center, Huntress's own write-up, Sophos, CISA, NIST, OWASP and Microsoft Learn on November 12, 2026. ConnectWise says the initial access method and full scope are still under investigation, so this post doesn't guess at them.

What happened on November 11

Here is what the primary sources say, and how sure each one is.

Huntress (November 11, updated 11:34 PM ET):

  • Starting around 17:47 UTC (12:47 PM ET), Huntress saw an operator use the Run command feature on multiple cloud-hosted ScreenConnect panels to push commands to groups of managed endpoints, running as SYSTEM.

  • The commands installed an attacker-controlled ScreenConnect client and a toolkit Huntress calls SCKit, a 16-file package of scripts and configuration that sets up persistent access and a "relay trap" for intercepting session cookies.

  • Huntress counted at least 288 hosts across 48 organizations among its own customers, all hit within about 90 minutes, roughly a third of them in the first 10 minutes. That is what Huntress could see, not necessarily the total.

  • In its update, Huntress said the activity "seems to be related" to RMM+. It wrote that RMM+ "bypasses ScreenConnect's user accounts, roles, and MFA, and authenticates requests with a shared access key stored in the extension's settings," and that in versions 1.0.31 and 1.0.32 the key defaulted to blank and the check didn't reject an empty value.

  • Huntress also said most RMM+ installs it found "were legitimate, usually added for a project and left in place."

  • SCKit contains code designed to jump from an infected endpoint to a technician who connects to it, then on to other machines that technician's panel can reach. Huntress says that console-takeover attempt failed in every recovered host log it has checked so far. That's why both Huntress and ConnectWise warn against triaging suspect machines over ScreenConnect itself.

  • Huntress noted that the current ScreenConnect releases, 26.6.6.9747 and 26.7.6.9783, were among those affected. Read that as "being fully patched didn't stop it," not as a newly found ScreenConnect flaw.

ConnectWise (advisory first posted 22:45 UTC on November 11, updated 02:45 UTC on November 12):

  • ConnectWise says it has found "no evidence of a compromise of ConnectWise infrastructure or a vulnerability in ScreenConnect itself."

  • Its initial assessment is that the activity "may involve a threat actor's abuse of independent third-party extensions installed in partner environments," and that it is still validating that assessment. The initial access method and full scope remain under investigation.

  • It identified approximately 400 partner environments with at least one of five related extensions installed (Labtech Plus, RMM Plus, RMM+, RMM Plus Password Link and RMM+ Password Link) and disabled them wherever it could. It stresses that having one installed "does not establish compromise."

  • It describes these extensions as providing "an API capable of executing remote commands," protected "by a passphrase configured by the partner during setup."

  • On-premises partners have to disable the extensions themselves. ConnectWise says not to re-enable or reinstall them pending further guidance.

Sophos separately reported malicious ScreenConnect deployments in the same campaign, coming from three ScreenConnect instances it believes were compromised, and recommends reviewing command history for unexpected client deployments.

We haven't found a public statement from the extension's developer, and no fixed version has been announced. Until ConnectWise says otherwise, the advice is disable, not "update."

"MFA didn't fail. It was never asked."

To see why strong technician logins didn't help, think of ScreenConnect as a building with two doors.

The front door is the console. A technician opens the web panel, types a password and approves an MFA prompt. ScreenConnect then checks their role and session groups to decide which machines they can reach and what they can do. That's the path all the usual advice protects: unique passwords, MFA, least-privilege roles. ConnectWise's own guidance on securing user accounts covers it well, and cloud instances turn on two-factor authentication for internal users automatically.

The side door is the extension's API. Integrations exist so another system, such as an RMM or a documentation tool, can tell ScreenConnect what to do without a person clicking through the console. Huntress describes RMM+ as letting "another tool manage ScreenConnect sessions without a technician working through the console." That other tool doesn't have a phone to approve an MFA prompt. So the extension does its own check: does the request carry the right access key?

That check is the entire lock on the side door. There's no user, no role, no MFA prompt and no login event. Once the key matches, the request is trusted. In RMM+ 1.0.31 and 1.0.32, according to Huntress, a key that was never set matched a request that sent no key.

Three things follow from that, and they apply far beyond ScreenConnect:

  1. MFA protects people, not keys. MFA is a control on the human sign-in path. A request that authenticates with a key never touches that path, so the strength of your MFA is irrelevant to it.

  2. Roles don't apply either. Huntress says the extension bypassed ScreenConnect's roles. Whatever the extension could do, a key holder could do, across every machine the panel reached.

  3. Your login logs stay quiet. If you only watch for strange sign-ins, a key-based attack looks like nothing happened. Huntress reported "no corresponding login activity." The evidence lives in command history, Toolbox history and extension logs, not the sign-in report.

This isn't a ScreenConnect-only problem. It's the same pattern we covered for Microsoft 365 in MFA can't stop attackers logging in as your apps: when the "user" is software, its secret is the whole identity.

An integration key is a credential, not a setting

The RMM+ settings page reportedly described the access key this way: "This is like a password. This is required to be sent in the POST request to access the functions. Set to a strong key." It was right, and that's the lesson. A field that sits among dozens of extension settings is still, functionally, an admin password for every endpoint the tool can reach.

The standards bodies have been saying this for years.

OWASP lists "API keys" plainly in its API Security Top 10: "OAuth is not authentication, and neither are API keys." It also says API keys should only authenticate API clients, never stand in for user authentication. The OWASP Non-Human Identities Top 10 (2025) reads like a description of this incident:

  • NHI1, Improper Offboarding: access keys left in place after they're no longer needed. "Added for a project and left in place" is exactly this.

  • NHI3, Vulnerable Third-Party NHI: if a third-party extension is compromised or vulnerable, attackers can misuse the permissions it was granted.

  • NHI4, Insecure Authentication: weak or outdated authentication methods for service-to-service access.

  • NHI5, Overprivileged NHI: an integration that can run commands on every endpoint when it only needed to read a session list.

  • NHI7, Long-Lived Secrets: keys that never expire give an attacker access without a time limit.

NIST SP 800-53 Rev. 5 treats this as basic hygiene. Its authenticator management control (IA-5) calls for changing default authenticators before first use, changing or refreshing authenticators on a schedule or after events such as suspected compromise, and protecting authenticator content from disclosure. A blank key is the ultimate unchanged default. The least-privilege control (AC-6) applies to integrations as much as people. And NIST SP 800-161 Rev. 1 frames third-party software components as supply-chain risk you have to manage, not trust by default.

CISA's Secure by Design alert on default passwords puts the burden on vendors: products shouldn't ship with universally shared default passwords, and setup should force an instance-unique credential. CISA notes that "relying upon thousands of customers to change their passwords is insufficient." A key that defaults to blank and still works is the same failure. Until every vendor gets that right, though, the customer still has to check.

Microsoft Learn says the same for its own platform. Its workload identity overview explains that a software workload "may deal with multiple credentials," and that "it's also hard to track when a workload identity is created or when it should be revoked." Its app registration best practices say to prefer managed identities or federated credentials over secrets, to "rollover credentials frequently and don't share credentials across applications," and to review app owners every few months. Different product, same lesson.

Put simply, every integration key needs four things:

  • An owner: a named person who knows why it exists and gets asked when it should go away.

  • A scope: it can do only what the integration needs, ideally read-only if that's enough.

  • A lifetime: it's rotated on a schedule and immediately after any staff change or incident.

  • Monitoring: someone can see what it does, separately from human sign-ins.

How this differs from the vendor-app breach

If you read our earlier post, The app you forgot you approved, this may sound familiar. It's a different failure, though, and the fix lives in a different place.

  • That incident was about vendor-owned apps inside your Microsoft 365 tenant. The vendor held the credential, and your job was to audit apps, permissions and sign-ins in Microsoft Entra.

  • This incident is about an extension the MSP installed in its own remote-access tool, protected by a key the MSP was supposed to set. Microsoft Entra logs won't show it. The fix sits in the ScreenConnect admin pages, the extension settings and the MSP's own processes.

The shared thread is that machine credentials outlive the projects that created them. The audit just has to happen in more than one console.

The MSP checklist for this week

If you run ScreenConnect, cloud or on-premises, work through these in order. ConnectWise's advisory is the authority here; follow it where it differs.

  1. Confirm the five extensions are disabled. In each instance, open Admin > Extensions and look for Labtech Plus, RMM Plus, RMM+, RMM Plus Password Link and RMM+ Password Link. ConnectWise disabled them on hosted instances where it could. On-premises, you must do it yourself. Don't re-enable or reinstall them until ConnectWise says it's safe.

  2. Preserve evidence before you clean anything. ConnectWise asks partners to keep authentication, extension installation, administrative, Run Command and Toolbox records. Export them before changing configuration.

  3. Review Run Command and Toolbox history. Look for mass commands nobody on your team ran, especially anything that downloads or silently installs a ScreenConnect client. Huntress says the November 11 activity looked like one operator sending commands to many endpoints at once.

  4. Look for unexpected ScreenConnect clients and sessions. A second ScreenConnect client service on a managed machine, or a session pointing at an instance that isn't yours, is a red flag. Huntress and ConnectWise both publish indicators to check against. Use their lists rather than anyone's summary.

  5. Investigate suspect machines out of band. Don't open an interactive ScreenConnect session to a machine you suspect. Use EDR live response or another separate tool. If a technician already connected, treat their workstation as possibly exposed and check it from a clean device.

  6. Reset access from a trusted device. ConnectWise recommends revoking active ScreenConnect and identity-provider sessions, resetting relevant credentials, enforcing MFA and removing unnecessary admin access. Because SCKit targets session cookies, signing out sessions matters as much as changing passwords. If those technicians also run Microsoft 365 admin accounts, see Revoke sessions isn't enough: what ends Microsoft 365 access.

  7. Inventory every other extension. Don't stop at the five. For each extension, write down who installed it, why, when it was last used, and whether it has a key or secret. Remove anything left over from a finished project.

  8. Set, strengthen and record every extension key. For extensions you keep, confirm the key isn't blank, generate a long random one, store it in your password vault or secrets manager, and record its owner and rotation date.

  9. Repeat for your other tools. Your PSA, RMM, documentation platform, backup console and billing integrations all hold API keys too. Same questions: who owns it, what can it do, when was it rotated, and who would notice if it were misused?

  10. Tell your clients. Clients will hear about this. A short, factual note saying whether you used the affected extensions, what you checked and what you changed builds more trust than silence.

For on-premises servers, a quick read-only look at the extensions folder can complement the Admin page. Mandiant's ScreenConnect hardening guide points to this folder. The folder names are IDs, not friendly names, so match them against the Admin > Extensions list:

# List extension folders on an on-premises ScreenConnect
# server. Read-only. Run in an elevated PowerShell window.
$root = 'C:\Program Files*\ScreenConnect\App_Extensions'
Get-ChildItem -Path $root -Directory |
  Select-Object Name, CreationTime, LastWriteTime |
  Sort-Object LastWriteTime -Descending

A folder with a recent write time that nobody remembers is worth a question.

If you're an MSP client: what to ask

Most small businesses never see their MSP's remote-access console, and that's fine. You can still ask good questions. You don't need to know what an extension is; you just need clear answers.

  1. "Do you use ConnectWise ScreenConnect, and did you have any of the five extensions ConnectWise disabled on November 11?" A yes isn't automatically bad news. ConnectWise says having one installed doesn't prove compromise. What matters is what they checked.

  2. "Did you review command history for our machines, and did you find anything you didn't run?" You want a specific answer with a date.

  3. "Have you checked our computers for an extra remote-access client?" This is the main thing the attackers installed.

  4. "Which third-party add-ons can run commands on our machines, and who owns each one's key?" If the answer is "we'd have to check," ask them to check and report back.

  5. "How often do you rotate integration keys, and what happens when a technician leaves?"

  6. "Would you tell us within a set time if a tool you use to manage us is involved in an incident?" If your contract doesn't say, this is a good time to add it.

Ask politely. Your MSP had a long night too. The goal is evidence, not blame.

Integration-key hygiene beyond this incident

The RMM+ flaw was specific. The habits that would have limited it aren't. Here's the short version for any tool that talks to another tool.

  • Never accept an empty secret. If you build or buy integrations, make sure a blank or default key fails closed. CISA's guidance is clear that setup should force a unique credential.

  • Prefer keyless where the platform allows it. On Microsoft Entra, Microsoft recommends managed identities or workload identity federation over stored secrets, and certificates over passwords when you must store something. Fewer stored keys means fewer keys to leak.

  • One key per integration. OWASP's NHI9 warns against reusing one credential across services. A shared key turns one leak into many.

  • Least privilege for software too. If an integration only needs to read session lists, don't give it command execution. The same thinking applies to Microsoft 365 apps; see Stop giving apps Application.ReadWrite.All.

  • Store keys in a vault, not a settings page alone. Microsoft's Key Vault guidance and OWASP's Secrets Management Cheat Sheet both stress central storage, access control on who can read the secret, rotation and auditing.

  • Watch what software does, not just who signs in. Command history, API logs and service-principal sign-ins are where integration abuse shows up.

  • Offboard keys like people. When a project ends or a technician leaves, the keys they set up should be reviewed and rotated or removed.

CISA and partner agencies warned as far back as January 2023 (advisory AA23-025A) that attackers abuse legitimate remote monitoring and management software because it's trusted and often allowed through security tools. That warning still holds, and integrations are part of that trusted surface.

The tradeoffs

None of this is free, and it helps to be honest about the costs.

  • Disabling an extension breaks something. RMM+ was installed because someone needed another tool to manage sessions. Expect some automation to stop until there's a vetted replacement. That's an acceptable cost right now.

  • Rotation causes outages if nobody owns the key. Rotate a key without knowing what uses it and an integration fails at 2 AM. That's why owner and purpose come before rotation.

  • Fewer integrations means more manual work. Every removed add-on may mean a technician doing a task by hand. Keep the ones that earn their place and drop the ones nobody can explain.

  • Logging costs money and attention. Forwarding ScreenConnect and API logs to a SIEM isn't free, but it's the only place key-based activity shows up.

Key takeaways

  • The commands came in without a login. Huntress saw commands reach endpoints with no corresponding login activity and linked them to the RMM+ extension.

  • RMM+ authenticated with a shared access key that bypassed accounts, roles and MFA, and versions 1.0.31 and 1.0.32 accepted a blank key when none had been set, according to Huntress.

  • ConnectWise has not confirmed the initial access vector. It says it found no evidence of a ScreenConnect vulnerability or a ConnectWise infrastructure compromise, and it disabled five related extensions in about 400 partner environments as a precaution.

  • MFA protects human sign-ins. Anything that authenticates with a key skips that path entirely, so the key is the whole lock.

  • Integration keys are credentials. Each needs an owner, a narrow scope, a rotation schedule and its own monitoring.

  • Disable, don't update, the five named extensions until ConnectWise says otherwise, and investigate suspect machines out of band, not over ScreenConnect.

  • MSP clients should ask specific questions about extensions, command history and extra remote-access clients, and expect specific answers.

Official sources

About IAMPathway

IAMPathway publishes practical guides on Identity & Access Management: Microsoft Entra ID, governance, identity threats, Application Governance and AI Governance. Browse all topics or read the glossary.