N-able N-central – God Mode on the Management Plane, and the Hotfix That Wasn't Enough
24 minAudio
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! At the start of August 2026, attackers exploited an authentication bypass in N-able's N-central remote monitoring and management platform to obtain full administrative control of the console - and then used the product's own legitimate remote access feature to reach the machines it manages. The vendor shipped a fix. It was not enough. The six-day timeline. 1 Aug: N-able detects active exploitation of CVE-2026-18556 (CVSS 7.4 under v3.1, 8.2 under v4.0); all versions affected, hosted and on-premises. 2 Aug: hotfix 1, build 2026.3.1.7, plus a second advisory for CVE-2026-18577 - the residual bypass left by the incomplete fix (8.1 under v3.1, 8.2 under v4.0). 3 Aug: CISA adds 18577 to the Known Exploited Vulnerabilities catalogue. 4 Aug: CISA adds 18556; N-able confirms attackers obtained administrative access. 6 Aug: hotfix 2, build 2026.3.1.10 - required even if hotfix 1 was already applied. The patch-rate gap. Huntress observed 55.6% of reachable cloud N-central servers unpatched early on 3 August, falling to 13.6% overall by that afternoon - but 28.6% of reachable self-hosted servers were still unpatched. Hosted customers were fixed by the vendor; self-hosted ones had to fix themselves. The attack chain. Unauthenticated bypass to full admin ("god mode"), then abuse of the built-in Take Control remote access feature - with sessions logged under the default legitimate "MSP Support" account, so the attack looks like ordinary support work. Then domain controller reconnaissance, and persistence via a Cloudflare Tunnel (cloudflared) registered as a Windows service plus a suspicious svchost.exe in a user's Documents folder. No attribution - at all. Not formal, not suspected, not claimed. All ten published indicator IPs are Mullvad or NordVPN exit nodes. No ransomware was observed and no extortion claims have surfaced. Hunting artefacts: ui_access_control.log for the indicator IPs and the mspsupport identity; the Take Control logs under ProgramData; Windows event IDs 4102, 8192 and 8193; and any cloudflared service you did not install. Five durable lessons: give the management plane identity-provider-grade controls, including IP allow-listing and MFA on every operator account; recognise that the on-premises appliance is usually the one machine in your estate without EDR; understand that applying the vendor patch is not the same as being safe; treat remote control sessions as security events and review the out-of-hours ones weekly; and if you buy IT services, ask your provider precisely when they applied the August hotfixes. Because when a database is breached you lose data. When your management plane is breached, you lose the ability to trust anything else you are looking at - including your logs and your patch reports. Visit www.kinsoft.com.au to talk through your security and IT needs. Sources: N-able security advisories for CVE-2026-18556 and CVE-2026-18577 (1-6 August 2026); Huntress, "N-able vulnerability exploitation" (3 August 2026); BleepingComputer; Rapid7 Emergent Threat Response for CVE-2026-18577; Horizon3.ai attack research; CISA Known Exploited Vulnerabilities catalogue additions, 3 and 4 August 2026; The Hacker News; The Register.
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
When you think about the absolute most heavily guarded piece of software in any business, your mind naturally, you know, it drifts toward a few obvious suspects. Right. Like the giant edge firewall. Exactly. The firewall absorbing like millions of malicious pings from the public Internet. Or maybe you think of the central identity provider vaulting everyone's passwords or the backup platform holding the company's data in a digital safe. And it's entirely logical that we default to those systems. I mean, they are the traditional
fortresses of enterprise security. We are trained to view them as the primary targets because, well, for a very long time they were. But when you actually break down the architecture of a modern IT environment, the most powerful piece of software isn't any of those things. It is the remote monitoring and management tool, the RMM. Oh, yeah. The ultimate target. And honestly, it is the single platform that keeps IT administrators staring at the ceiling at two in the morning. So today we are taking a deep dive into the early August 2026 attacks on Enable's nCentral platform.
Which was just a wild situation. It really was. And to do this, we are pulling from a heavy stack of sources. We've got threat intelligence reports from Huntress, telemetry data, the official vendor bulletins from Enable, and advisories from CISA. Yeah. But our mission here isn't just to recap a news event or read a security bulletin to you. We are going to use this specific incident as a lens to understand exactly how modern attackers operate, why our conventional defenses are completely failing to stop them, and how the fundamental
concept of moving through a network has changed. OK, let's unpack this. If your company's network is a massive, highly secure office building, the RMM isn't just the master key. It's the security guard desk, the intercom system, and the building's ventilation system all rolled into one central console. That is a great way to put it. Right. Because it relies on an agent installed on every single server and workstation. It runs as system, which is the highest possible level of privilege on a machine. It executes code silently in the
background. And by design, it completely ignores internal network segmentation because its entire job is to reach out and touch every single device. That ventilation system. analogy really highlights the true scope of the risk for anyone listening to this. I mean, if you want to distribute a gas throughout an entire skyscraper instantaneously, you do not walk from room to room with a spray can. No, you put it in the vents. Exactly. The legacy model of hacking was about gaining a tiny foothold, you know, tricking an employee in marketing into clicking a phishing link, for instance.
Right. Then the attacker had to painstakingly hack their way across the network, server by server, trying to escalate privileges without tripping any internal alarms. That process takes weeks, sometimes months. But if an attacker compromises your RMM console, they completely bypass that entire phase of the operation. They're going to skip it. Totally. They do not need to figure out how to move laterally across your environment. They already are lateral. They've just seized control of a vendor -built, fully documented, and fully supported mechanism to run any code they want on any machine they want at any time.
Which takes us right into the events of early August. To grasp the severity of this exploit, we need to look at how the discovery unfolded. The core issue was a vulnerability labeled CVE -2026 -100556. And just to ground that acronym for a second, a CVE is a Common Vulnerabilities and Exposures Record. It's essentially a standardized ID tag the industry uses to track a specific software flaw. Right. Thank you. And this particular flaw was an authentication bypass. But the technical
flaw itself isn't actually the most alarming part of this story. The pacing of the discovery is what makes this a nightmare scenario. Yeah, because the context of the discovery dictates the entire defensive response. When White Hat security researcher finds a bug during a routine code audit, they usually follow responsible disclosure. Which is nice and quiet, right? Exactly. They privately notify the vendor, and the vendor gets a grace period, often 30 to 90 days, to quietly write, test, and distribute a patch before the public ever knows the flaw exists. that is an orderly process but that is not what happened
here no not at all Right. Our threat intel sources note this wasn't found in a lab. On Saturday, August 1st, Enable published their advisory because they detected active exploitation in the wild. Someone was already using this bypass against live targets. Which is terrifying. Yeah. And the flaw affected every version of nCentral, whether it was hosted by the vendor in the cloud or sitting on a physical server in a company's basement. And finding a flaw because it is actively being used against you means you're automatically losing the race. I mean, the building is active.
burning while you are still trying to sketch the blueprints for a fire extinguisher. Wow. Yeah. The vendor is forced to write code under immense pressure and network defenders are forced to apply updates without their usual testing cycles. And we saw the messy reality of that pressure immediately. The very next day, Sunday, August 2nd, the vendor ships an emergency fix, Hotfix 1. But simultaneously, they have to publish a second advisory for a completely new vulnerability ID, CVE -2026 -15777. All right, because the
fix was incomplete. They boarded up the front door, but they left the ground floor window wide open. The initial Hotfix addressed the specific path the attackers were using to bypass the authentication check. But there was a residual path in the code. that accomplished the exact same thing. And because the attackers were already inside the perimeter, actively poking at these systems, they identified that alternate path instantly. Right. They didn't even have to look hard. Over the next few days, CISA adds both of these flaws to their known exploited vulnerabilities catalog, which is basically
the government's official drop everything and fix this list. Finally, by Thursday, August 6th, Enable releases Hotfix2, which fully closes the window. Finally. Yeah. But amidst this six -day scramble, I noticed a huge debate erupting in the security community regarding the CVSS severity scores. Under version 3 .1 of the scoring system, this was rated a 7 .4 out of 10. But under the newer version 4 .0 framework, it scored an 8 .2. My immediate reaction to reading that debate in our sources was just confusion. Right, I remember
that. Aren't we missing the bigger picture by arguing over the underlying math of a severity score when servers are actively being compromised? What's fascinating here is that debate perfectly illustrates a chronic issue in the security industry, getting paralyzed by your own bureaucracy. Totally. The CVSS framework is just a mathematical formula used to prioritize patching, based on factors like how much privilege is required to execute an attack. Arguing over a 7 .4 versus an 8 .2 during an active incident is a complete distraction.
Yeah, it doesn't matter. The only two questions a defender should ever ask when an advisory drops are, is it being exploited right now? And what level of access does it give the attacker? In this case, the answers were yes, right this second, and total systemic control. The arithmetic of the score is irrelevant when the reality is absolute compromise. And the telemetry data from Huntress really hammers home how vulnerable organizations were during that week. They tracked Internet -reachable cloud servers, and by early Monday morning, just hours after that first incomplete hot fix dropped, 55 .6 % of those servers were
still completely unpatched. I mean, over half the Internet -facing attack surface was wide open. It's crazy, and... While that overall number dropped to 13 .6 % by Monday afternoon, digging into the telemetry reveals a massive, terrifying gap. The vendor -hosted clients saw their numbers plummet because Enable was forcing the fixes on the cloud side. Right, they had control there. But for the self -hosted on -premises environments, the companies running their own dedicated servers, 28 .6 % of them were still completely exposed.
That gap between cloud and self -hosted environments highlights the immense friction of on -premises management. If you manage your own nCentral server, you are solely responsible for noticing an urgent vendor advisory on a Saturday morning. Which, who is checking that on a Saturday? Exactly. You have to validate the patch, communicate the risk to your stakeholders, schedule the downtime, back up the server, and manually apply the update. That entire process takes time. And time is the one resource the attackers explicitly did not give them. Not at all. Leaving nearly 30 % of
self -hosted critical infrastructure exposed days into an active automated attack campaign is a catastrophic failure of agility. Because that patch took days for self -hosted administrators to apply, it created a massive window of opportunity. But what fascinates me isn't just that the attackers walked through that window. It is how they completely changed their disguise the second they crossed the threshold. Oh, this is the best part. Let's look at the actual playbook they used. First, they leveraged the unauthenticated bypass to
jump straight to full administrator access on the console. The threat researchers called this God Mode. Which feels entirely appropriate. It's very appropriate. They didn't steal a password or hijack a multi -factor authentication token. They sent a maliciously crafted web request and the server effectively just rolled out the red carpet. The system logic was fundamentally broken, granting them the highest level of authority without any identity verification at all. But the truly evolutionary aspect of this campaign is what happened next. Right. Here's where it
gets really interesting, because the logical next step for a traditional attacker would be dropping a stealthy remote access Trojan or, you know, deploying a custom piece of malware to solidify their grip on the network. But they didn't do that. Instead, they utilized Ncentral's own internal take control feature. Yeah. And take control is the built in legitimate remote desktop protocol. If you are sitting at your desk and your screen freezes and you call your IT help desk, they will use a tool just like
Take Control to view your screen and move your mouse. It is a digitally signed, vendor -supplied executable. It is natively installed on every single endpoint in the corporate network. Because of that pedigree, the local antivirus software and the advanced endpoint detection and response tools implicitly trust it. So they didn't break a window to gain access to the interior offices. They just walked into the lobby, put on a perfectly tailored security guard uniform, and grabbed a clipboard. That's exactly what they did. And we know this because of the audit logs. The attacker's
remote sessions were logged under the username MSPSupport. That happens to be the default legitimate account name for the take control feature. If an IT analyst was casually reviewing a dashboard, it would look exactly like normal everyday support activity. The difficulty of detecting that kind of movement really cannot be overstated. If a threat actor is utilizing the exact same remote access tool under the exact same service account name and writing the exact same encrypted network protocols as your legitimate IT staff, the standard
technical tripwires will never trigger. They're just blind to it. Completely. This is the textbook definition of living off the land. The attackers are weaponizing the environment's own native administrative tools against it. Unless a human defender is actively correlating remote control sessions with expected business hours and known support tickets, this activity is functionally invisible to an automated defense grid. So they are effectively invisible, and they have the keys to the entire kingdom. Once they established that foothold, the logs show they immediately targeted the domain controllers, you know, the
servers that access the master address book and permission grantors for the entire Windows network. The big targets. Yeah. But they also needed to ensure they could maintain access even if a frantic IT administrator finally applied the enable hotfixes and closed the initial front door. To establish that persistence, they used a combination of highly sophisticated routing and incredibly rudimentary psychological tricks. The persistence mechanisms are a fascinating study in contrast. For the advanced route, they deployed a Cloudflare tunnel,
specifically the Cloudflare binary, and registered it to run silently as a background Windows service. Wait, a Cloudflare tunnel? I mean, I use Cloudflare for basic web... post and content delivery, why would an advanced threat actor use a commercial web performance tool as a network backdoor? Because it provides brilliant network camouflage. A Cloudflare tunnel is designed to create an outbound -only encrypted connection from a server directly to Cloudflare's global edge network. Most mature corporate environments utilize egress filtering,
meaning the firewall actively blocks internal servers from initiating traffic out to the internet unless it is going to a pre -approved destination. But very few organizations block traffic destined for Cloudflare because a massive percentage of the legitimate Internet relies on their infrastructure. Oh, that makes sense. When the corporate firewall sees encrypted traffic initiating outward to a known Cloudflare IP address, it waves it right through, assuming it is routine web traffic. By doing this, the attacker establishes a permanent, heavily encrypted reverse shell backdoor that
effortlessly glides right past traditional perimeter defenses. Man, that is painfully clever. They are hiding their malicious traffic inside the sheer volume of normal web traffic. Exactly. But then they paired that advanced encrypted tunnel with a trick so old it practically has dust on it. They hid a malicious executable file right inside a user's documents folder, and they named the file svcoast .exe. And that tactic proves that human psychology remains one of our most exploitable vulnerabilities. Seriously. svcoast .exe stands for service host. It is a
core Windows system process used to group various dynamic link libraries. If you are listening to this right now on a Windows machine and you open your task manager, you will likely see dozens of svcoast .exe processes running simultaneously. Yeah, it's everywhere. Our brains have been conditioned over decades of IT work to filter that specific file name out as background noise. When an exhausted security analyst is scrolling through a massive directory listing during a high -pressure investigation, their eyes will literally glaze over that word.
It's like seeing a stop sign on your commute. You just stop registering it. Precisely. Logically, a core system file has absolutely no business running out of a user's documents folder. But the cognitive blind spot is so deeply ingrained that the attacker bets the analyst will just keep scrolling. And they're usually right. Wow. With an attack sequence this calculated from the zero -day exploit to the outbound encrypted tunnels and the psychological camouflage, The natural question our sources try to answer is, who orchestrated this? And the answer is a massive
void. There are zero formal attributions. No nation -state hacking group has claimed responsibility. They didn't drop a ransomware payload to extort a company. And no ransom demands have surfaced on the dark web. The major cybersecurity agencies and forensic firms are completely silent on who pulled the trigger. The lack of a clear motive or calling card is deeply unsettling. Usually the infrastructure an attacker leaves behind gives us clues. But the only breadcrumbs published in the threat intelligence reports are 10 indicator
IP addresses. Right. And every single one of those addresses belong to a Malvad or NordVPN exit node. Which means what exactly for the investigation? It demonstrates an incredibly high level of operational security. If an adversary launches an attack from a compromised server in a specific geographic region. Investigators can often glean clues from the hosting provider or the regional routing data. Sure. By deliberately anonymizing all of their command and control traffic through commercial, strict, no -log VPN services, the attackers are
aggressively scrubbing their digital fingerprints. They are categorically ensured they cannot be traced back to a physical location or a specific state apparatus. If we step back from the specific technical indicators of this N -able incident, it becomes clear we're looking at a much broader structural reality. Because N -able is just the current headline. If you trace the lineage in our sources exactly one year prior to this, two other nCentral zero -day vulnerabilities were exploited in a very similar fashion. Right. And before that, the industry was rocked by compromises
involving Kaseya VSA, ConnectWise Screen Connect, SimpleHelp, and famously, the SolarWinds Orion breach. If we connect this to the bigger picture, the pattern is undeniable. This is not a critique of one specific vendor's engineering team. It is a fundamental law of leverage in the digital age. Exactly. If you are a highly capable threat actor, it makes absolutely zero economic sense to spend months hacking 50 individual companies one by one. You target the single management
platform those 50 companies use to administer their networks. Work smarter, not harder. Yep. Management platforms are the absolute highest leverage targets on the board. If you rely on one of these platforms, which almost every modern organization does, you can no longer operate under the assumption that it might eventually be targeted. You must operate under the absolute certainty that it is currently under a sophisticated, continuous siege. So if we accept that they are coming for the RMM, how do we actually defend it? Knowing the threat exists only matters if
we change our behavior. Based on the post -incident reports, there are several concrete, durable takeaways for anyone listening who manages an IT environment or hires a provider to do it for them. The first is a complete reevaluation of how we protect the management plane. Historically, we have placed massive friction in front of our email and identity systems. We mandate complex conditional access policies and multi -factor authentication for a simple Microsoft 365 inbox. Oh yeah, MFA on everything. Yet the RMM console,
which has the power to systematically wipe every server in the company, often sits exposed to the public internet, protected by nothing more than a static password. That architecture is no longer viable. Not at all. The RMM console must be treated with the exact same paranoia as a core identity provider. Phishing -resistant MFA is non -negotiable for every single account, especially the automated service accounts that administrators just forget about. Furthermore, the login portal should never be indexed on the public web. It must sit behind a strict IP allow list, ensuring that only physical office locations
or authorized corporate VPNs can even load the login page. The second major takeaway addresses a highly specific technical challenge highlighted in this incident, the appliance blind spot. Many organizations deploy these tools as prepackaged, on -premises virtual appliances the n central server in this attack was running on a custom alma linux 9 operating system provided by the vendor and an appliance is essentially a black box you do not control the underlying operating system because of that you usually cannot install
your standard endpoint detection and response software inside the appliance because it just rejects it right yeah the custom linux kernel will either reject the security agent or the vendor will explicitly tell you that installing third -party software voids your support contract this creates a deeply ironic and dangerous situation The single most privileged, powerful machine in your entire network architecture is the only machine completely devoid of internal security monitoring. Which is terrifying. So if you are legally or technically barred from putting a
security agent inside the box, how do you monitor it? You have to compensate by monitoring the perimeter of the box with intense paranoid scrutiny. You implement deep network flow analysis and packet capture directly outside the appliance's virtual network interface. You configure the appliance to forward every single internal audit log it generates to a completely separate immutable server. Most importantly, you create hard immediate alerts for network anomalies. The RMM appliance should only ever communicate with your internal
endpoints and the vendor's update servers. If that black box suddenly initiates an outbound connection to an unknown server in Eastern Europe or attempts to negotiate a Cloudflare tunnel, you need alarm bells ringing in the Security Operations Center instantly. That makes total sense. And that brings us to a massive cultural shift regarding how we view software updates. The timeline of this incident proves that patching is a continuous process, not a discrete event. Applying Hotfix 1 on Sunday was not the finish line. The IT industry is culturally conditioned to view patching as a checklist item. A critical
vulnerability drops. The infrastructure team ruins their weekend to apply the patch. They verify the server reboot successfully. They close the help desk picket. And they go to sleep feeling accomplished and secure. But we applied the patch does not mean we are secure. Because if you close the ticket on Sunday, August 2nd, and stopped paying attention, you were completely exposed to the residual bypass until Thursday, August 6th. Exactly. The workflow has to evolve. Applying the initial hotfix is only step one. Step two
requires dedicating personnel to actively monitor that specific vendor advisory day after day. Right. You have to refresh the threat intelligence feeds looking for the inevitable update that says the previous patch was bypassed by a new method. Please apply hotfix two immediately. An incident response cycle is not over when the server reboots. It is over when the vendor advisory definitively stops evolving. We also have to address the invisibility of the attack itself. Since the adversaries are living off the land and using legitimate tools like take control, traditional antivirus isn't going to save you.
The countermeasure here requires behavioral logging. Yeah, if the malicious activity is just a good tool being used at a very bad time, your only defense is behavioral context. Organizations must extract the logs of every single remote control session, whether you use Take Control, AnyDesk, Screen Connect, or TeamViewer, and centralize them in a dashboard. And actually look at them. Yes. Centralizing them isn't enough. A human being must actively review them. A disciplined five -minute weekly review of all remote sessions
that occurred outside of normal business hours would have illuminated this attack immediately. If an IT manager sees the MSP support account establishing a remote desktop session to a domain controller at 3 .00 a .m. on a Sunday, they instantly know they have an active intruder. Finally, there is a crucial takeaway for the business leaders listening who outsource their technology to a managed service provider. When you meet with your MSP, asking them a vague question like, are we secure? is entirely useless. The sales team is always going to say yes. You need to
test their operational reflexes. You cut through the marketing language by asking highly specific historical questions. You sit down with your account manager and ask, during the first week of August 2026, exactly how quickly did your team apply both enable hot fixes to our management infrastructure? Oh, that's a great question. If they can look at their records and reply with a specific timeline, saying they applied hot fix one on Sunday night and hot fix two on Thursday morning, that demonstrates a mature functioning patch management pipeline. But if they stutter,
if they have to go ask an engineer and get back to you, or if they don't even know what you were talking about, it reveals that they are flying completely blind. And if they use a different platform, you simply substitute the name of their tool and the date of its last major zero day. Yes, it is the ultimate stress test for an IT partnership. It forces transparency because at the end of the day, there is a fundamental existential difference between a standard corporate data breach and a breach of an RMM platform. When a database is compromised, you lose data. It
is a terrible, expensive day, but the damage is a known quantity. But when your management plane is breached, the very tool you rely on to observe and control your network has been weaponized against you. You lose the ability to trust the dashboard you are looking at. And that is the most unsettling reality of modern infrastructure defense. It leaves us with a deeply uncomfortable scenario that every network defender eventually has to grapple with. If a highly disciplined attacker silently compromises the exact software you rely on to manage your environment, and they
have the privileges to manipulate the very audit logs you use to monitor it, how do you even begin an incident investigation when your own digital environment is actively lying to you? It is the equivalent of discovering the security guard you hired to watch surveillance cameras is the one looping the video feeds. It forces you to question the foundational reality of your entire network. Take a hard look at your own management systems with fresh, deeply questioning eyes this week.
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Japanese home goods company Iris Ohyama (online store operator Iris Plaza) found a suspicious file in an external...
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Australian online retailer Who Gives A Crap told customers on 7 November 2026 that their information was accessed...
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Last week in tech, 7 to 14 November 2026. Microsoft's November Patch Tuesday (10 November) fixes well over a...
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Chainalysis disclosed on 30 October 2026 that an external actor using "offensive agentic techniques" accessed...
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! NSW Health reported unauthorised access to its Oral Health Fee For Service Scheme website on 29 October 2026;...