Phishing abuses RMM tools for persistent access by tricking one employee into installing legitimate, digitally signed remote management software, which then hands an attacker a quiet, trusted way back into the network whenever they want. Microsoft Defender Experts documented this pattern in a September 2026 report: lures dressed up as Zoom installers, PDF updates, e-cards and delivery notices delivered a signed MSP360 agent, and that agent then pulled down ConnectWise ScreenConnect as a second remote channel. Nothing in the chain needed classic malware. The lesson for a 50 to 500 person business is blunt. You will not win this by training people to spot every lure. You win it by knowing, in writing, which remote tools are allowed on your machines, and by treating any other one as an incident. Below are the five checks our team runs for clients, and why pulling one rogue agent is rarely the end of it.
Overview: Five Things to Take From This Campaign
Here is the short version for an operations director or CIO who has five minutes before the next meeting.
- The tool is real, so antivirus stays quiet. The installer was a legitimately signed remote management product. Signature-based tools have no reason to flag software that IT teams install every day.
- Attackers now plant two doors, not one. The first agent installed a second remote tool. If you find and remove one, the other still works.
- Elevation was the choke point. Per Microsoft, the installer needed administrator approval, and installs where elevation was denied did not complete. Local admin rights decide whether this attack lands.
- An approved list beats a blocked list. New remote tools appear constantly. Blocking names one at a time loses; allowing only your sanctioned tool wins.
- Cleanup means credentials too. Any account that installed an unapproved agent should be treated as compromised, with its password reset and its activity reviewed.
These five points shape the rest of this article, and each check below maps to one of them.
Why Phishing That Plants RMM Tools Slips Past Normal Defenses
Rogue RMM installs slip past normal defenses because every component in the chain is a trusted, signed business application doing exactly what it was built to do. Remote monitoring and management software, the same category your IT provider uses to patch and support your laptops, is designed to run with high privileges, start at boot, survive reboots and accept inbound connections. Those are precisely the traits an intruder wants.
Microsoft’s write-up describes the persistence in plain terms. The installer registered two Windows services, added registry Run entries so tray apps launch at sign-in, and opened an inbound Windows Firewall rule for the agent. To an endpoint tool, that looks like a normal software install. To an attacker, it is a foothold that outlives a password change.
We see the same blind spot in client environments. A help desk technician sees “remote support agent” in a process list and assumes it belongs to the IT provider. Nobody checks which vendor, which tenant, or who approved it.
Does user awareness training stop this kind of attack?
User awareness training lowers the click rate, and that matters, because fewer clicks means fewer chances for the installer to run. Our team still recommends it, and the lure themes in this campaign (meeting invites, document signatures, package delivery) are exactly the ones worth drilling. Our guide to recent phishing attack trends covers how those lures change shape.
The opposing view is just as fair. Microsoft observed many differently named files containing the same installer, hosted on attacker domains, compromised websites and legitimate cloud storage. Lures are cheap to rotate, and a busy finance clerk expecting a signed contract will click eventually. Training reduces exposure; it does not close the door.
Both views hold. Training is a filter, not a control. The control has to sit on the endpoint, where the installer either runs or does not.
Why does a signed installer look safe to security software?
A signed installer looks safe because code signing proves who published the file, not what the person running it intends. Signature checks answer “is this really from the vendor?” and the honest answer here was yes.
On the other side, modern endpoint detection does look at behavior, and a remote agent spawning PowerShell that downloads a second remote tool from an unfamiliar host is unusual behavior. Microsoft published hunting queries for exactly that sequence. So detection is possible, but only if someone is watching the alerts and knows which RMM is supposed to be there.
That is the practical gap. Behavior alerts without an approved-tool baseline produce noise that gets dismissed.
Is this only a problem for businesses that already use an RMM?
It affects every business, including those that run an RMM tool and those that do not. If you have no RMM at all, any remote agent appearing on a laptop is an obvious anomaly, which actually makes detection easier when someone is looking.
If you do run one, the risk is subtler. A second, unfamiliar agent blends in beside the legitimate one, and staff assume both belong to IT. Our experience is that businesses with an established RMM are more likely to miss a rogue one, not less, because the category already feels normal on their machines.
How the Two-Agent Pattern Changes Incident Response
The two-agent pattern changes incident response because finding and removing the first rogue tool no longer means the attacker has lost access. In this campaign the MSP360 agent ran PowerShell that silently installed a ScreenConnect client from attacker infrastructure. The actor then used ScreenConnect’s file transfer feature to stage credential and browser-password utilities, with filenames chosen to look like Windows, Defender and Phone Link components.
Think about what a typical small-business response looks like. Someone notices an unknown agent, uninstalls it, runs a scan, and closes the ticket. In this scenario, the attacker is still connected through the second tool, now with credentials harvested in the meantime.
That is the core of our view on this report. Removing one rogue agent is not cleanup. Cleanup is proving that every remote tool on every machine is one you approved, and that every account touched has been reset. Our piece on advanced persistent threats explains why intruders invest in redundancy like this.
Should you wipe a machine that had a rogue RMM agent installed?
Reimaging the machine is often the cleanest answer, because it removes services, scheduled tasks and registry entries you may never find by hand. For a single laptop with good backups, it is usually cheaper than an investigation.
The counterpoint is that wiping destroys evidence. If the agent ran under a system-level account or touched a file server, you need logs and timelines first, or you will not know which other machines and mailboxes to check. Microsoft specifically advises deeper investigation when a system-level account was used.
Our approach sits between the two: capture what matters, then reimage.
Do you need to reset passwords after removing the agent?
Yes. Microsoft’s guidance is to reset passwords for any account used to install an unapproved RMM tool, and the staged browser-password utilities in this campaign make the reason obvious. Saved browser credentials for banking, payroll and cloud apps may already be in the attacker’s hands.
Some teams worry that broad resets disrupt operations. That is a real cost, so scope resets to the affected user, any admin account that approved elevation, and any service whose password was saved in that browser. Then turn on sign-in alerts for those accounts.
How long can a rogue remote agent go unnoticed?
A rogue remote agent can go unnoticed for months, because it behaves like routine IT software and nobody is assigned to question it. Industry incident reports this year describe remote clients sitting undetected for long stretches before being used to change mailbox rules or move data.
The other side is that a single quarterly inventory catches most of them. Dwell time is long only where nobody is looking. Our related article on access permissions and insider risk covers the same visibility gap from the account side.
The 5 RMM Security Checks Every SMB Should Run
These five checks are the durable answer to RMM abuse, and none of them depends on recognizing the next phishing lure. We run them as part of our managed IT services, and any internal IT team can run them too.
- Write down your one sanctioned remote tool. Name the product, the vendor tenant or account it connects to, and who owns it. Everything else is unapproved by definition.
- Allow only that tool with application control. Use Windows Defender Application Control or AppLocker publisher rules to block unapproved IT management software. Microsoft also notes Defender for Endpoint can block specific signed applications by certificate.
- Remove local admin rights from daily accounts. The installer in this campaign needed elevation. Standard user accounts would have stopped it outright.
- Hunt for unapproved agents every quarter. Search services, installed software and outbound connections for any remote tool that is not on your list. Investigate any hit as an incident, not an IT chore.
- Put MFA on your real RMM console. If the attacker cannot hijack your legitimate tool, their only path is installing a new one, and checks two through four are built to catch that.
Is application allowlisting too heavy for a small business?
Allowlisting used to mean months of tuning, and for a 40-person firm with no IT staff that is a fair concern. Mistakes can block a line-of-business app on a Monday morning.
The narrower version is far lighter. You do not have to allowlist everything; you can start by blocking the remote management category except for your sanctioned vendor. That single rule addresses this whole class of attack, and it is the approach we start most clients on.
What should you ask your IT provider about RMM security?
Ask three direct questions: which remote tool do you use on our machines, how would you know if a different one appeared, and is MFA enforced on your console? A good provider answers all three without hesitation.
Some providers will say their tool is secure and stop there. That answers the wrong question. The risk in this report was never your provider’s agent being exploited; Microsoft saw no exploitation of ScreenConnect itself. The risk is a second agent your provider never installed.
Which logs prove an RMM agent was installed without approval?
Windows service creation events, software install logs and firewall rule changes are the clearest evidence. Microsoft’s report names the exact artifacts: two agent services, Run keys for tray apps, and an inbound firewall rule on a high UDP port.
The limitation is that small businesses often do not keep these logs long enough. Endpoint telemetry fed into a monitored platform solves that, which is where managed detection and response earns its cost.
How Phishing Lures for RMM Installs Keep Changing
Phishing lures for RMM installs keep changing because the payload stays constant while the wrapper is nearly free to rotate. Microsoft listed meeting requests, Zoom and Google Meet installer prompts, Adobe updates, RSVP e-cards, job offers, signature requests and fake package notices, all leading to the same installer. AI writing tools make each variant more convincing, as covered in our piece on AI-generated phishing threats.
This is why our recommendation weights the endpoint controls above lure recognition. Lure lists date quickly; an approved-tool policy does not. Our phishing attacks resource is a good primer for staff, and it works best paired with the five checks above.
Frequently Asked Questions
What is a rogue RMM agent?
A rogue RMM agent is legitimate remote monitoring and management software installed without your approval, usually through phishing. It gives an outsider the same remote control your IT provider has. Because the software is signed and common, it often goes unflagged.
How did the campaign Microsoft reported work?
Phishing emails led victims to download a signed MSP360 installer disguised as a meeting app, PDF update or similar file. Once running, it used PowerShell to install ScreenConnect as a second remote channel. The attacker then staged credential-theft tools through that second channel.
Can antivirus alone stop phishing that abuses RMM tools?
Antivirus alone usually cannot stop it, because the installers are legitimate signed software. Behavior-based detection and application control are the controls that work. An approved list of remote tools is what makes those alerts meaningful.
What is the fastest single fix for a small business?
Removing local administrator rights from everyday user accounts is the fastest single fix. Microsoft noted that installs failed where elevation was denied. Pair it with a written list of your one approved remote tool.
Does this mean RMM software is unsafe to use?
No, RMM software remains a sound way to support and patch machines. The danger is unapproved RMM tools, not the category itself. Enforce MFA on your sanctioned console and block every other remote tool.
Who Is Behind This Guidance
Our team has spent years managing remote access for small and mid-sized businesses, which means we live on both sides of this problem: we depend on RMM tools to support clients, and we hunt for the rogue ones attackers plant. The checks in this article come from that daily work, not from a vendor checklist. Your team does the hard part of running the business; our job is to make sure the remote tools on your machines are only the ones you chose. Mindcore is led by CEO Matt Rosenthal, who focuses on giving growing businesses enterprise-grade security practices at a size and cost that fits them, with practical controls over shelfware.
Get an Approved-Tool Baseline Before the Next Phish Lands
The takeaway from Microsoft’s report is that phishing abuses RMM tools for persistent access by exploiting trust, not technical flaws, and the defense that holds up is knowing exactly which remote tools belong on your machines. Spotting every lure is a losing game when wrappers rotate weekly and installers carry valid signatures. A written list of your one sanctioned tool, application control that enforces it, standard user accounts without local admin, a quarterly hunt for anything else, and MFA on your real console together close the door this campaign walked through. And if you do find a rogue agent, assume there is a second one and reset the credentials it could reach, because removal alone is not recovery. If you are not sure which remote tools are running across your laptops today, that uncertainty is the first finding. Our managed security services team can build that baseline with you and keep it current. Book a free strategy call and we will walk through your remote access inventory together.


