# Three more vulnerabilities went on the actively exploited list this month

> CVE-2026-18577 in N-able N-central joined CISA's exploited list alongside Cisco and Metabase flaws. All three are confirmed in use, not theoretical.

- Published: 2026-08-20
- Topic: [Security](https://futuretechnologyhq.com/topic/security/)
- URL: https://futuretechnologyhq.com/article/cisa-kev-august-2026-exploited-cves/
- Publisher: [Future Technology HQ](https://futuretechnologyhq.com/), free to read. Quote as "Future Technology HQ" with a link to the article.

## Key points

- CVE-2026-18577 in N-able N-central is an authentication bypass leading to account takeover, added to CISA's catalogue after real customer compromises
- Cisco Secure Firewall (CVE-2026-20349) and a Metabase SQL injection (CVE-2026-72898) went on the same list in the same week
- August Patch Tuesday cleared 421 CVEs including CVE-2026-68820, a Windows WinSock flaw already used for SYSTEM-level privilege escalation

## Article

CVE-2026-18577 is an authentication bypass in N-able N-central that leads to full account takeover. CISA added it to the Known Exploited Vulnerabilities catalogue after customer compromises, not after a researcher write-up.

N-central is remote monitoring and management software. It sits above hundreds of customer networks at once by design, holding the credentials and agent access needed to administer all of them. Compromise the RMM and you inherit everything underneath it.

## What the KEV catalogue actually signals

The distinction between this list and an ordinary advisory is the whole reason to prioritise it. CISA adds an entry only when there is reliable evidence of active exploitation in the wild, a CVE assigned, and clear remediation guidance available.

So the catalogue is not a severity ranking and not a list of theoretical risk. It is a record of flaws attackers are provably using. US federal agencies operate under a binding directive requiring remediation within a set window, and that deadline is the reason the list is a useful triage input for everyone else: it is the one feed where inclusion means somebody has already been attacked this way.

## The other CVE-2026-18577 neighbours on the CISA list

Two more went on in the same week. CVE-2026-20349 affects Cisco Secure Firewall. CVE-2026-72898 is a SQL injection in Metabase, the open source business intelligence tool that a lot of teams point directly at their production database.

Separately, August Patch Tuesday cleared 421 CVEs from Microsoft in a single release. Buried in that pile is CVE-2026-68820, a use after free in the Windows WinSock ancillary function driver, already being used in the wild to escalate to SYSTEM.

All four are confirmed exploited. That is the distinction worth holding onto when triaging: the KEV catalogue is not a list of things that could go wrong, it is a list of things that already have.

| CVE | Product | What it gives an attacker | Priority |
|---|---|---|---|
| CVE-2026-18577 | N-able N-central | Account takeover, then access to every managed estate | Today |
| CVE-2026-20349 | Cisco Secure Firewall | Perimeter foothold | Today |
| CVE-2026-72898 | Metabase | Whatever the connected database holds | Today |
| CVE-2026-68820 | Windows WinSock | Escalation to SYSTEM on an already-compromised host | This week |

The ordering reflects reachability rather than severity score. The first three are internet-facing by design and give an attacker something they did not already have. The fourth is a privilege escalation, which matters only once someone is inside, so it ranks below the three that let them in.

## The pattern is management tooling

Look at what these have in common. Remote monitoring software. A firewall appliance. A BI tool with database credentials. Attackers have stopped trying the front door and started going through the thing that watches the front door, because that thing has legitimate access everywhere and generates alerts nobody reads.

Three of the four, around 75% of this batch, are management or security infrastructure. That is not a coincidence of one week's reporting, it is the current shape of initial access.

It is the same shape as the [Defender zero day earlier this year](/article/shieldbreak-defender-zero-day-cve-2026-69414/) and the [hypervisor escape research that followed it](/article/kvm-januscape-hypervisor-vm-escape/). Security and management layers are high-value targets precisely because they were built to be trusted.

### RMM compromise is a force multiplier

The N-central entry deserves separate attention, because remote monitoring and management is the highest-leverage software category an attacker can reach.

An RMM platform exists to execute commands on managed endpoints, deploy software and run scripts, across every client an MSP serves. Its legitimate function is indistinguishable from an attacker's objective. There is no exploit needed after the initial access, because pushing a payload is simply using the product.

History makes the consequence concrete. A compromise of one RMM platform was used to deliver ransomware through roughly sixty managed service providers to around fifteen hundred downstream businesses simultaneously. A trojanised update to widely deployed monitoring software reached some eighteen thousand organisations. We covered a comparable case in the [SimpleHelp RMM zero-day rated CVSS 10.0](/simplehelp-rmm-zero-day/). If you are an MSP customer, your exposure includes your provider's tooling whether or not you have ever logged into it.

## Why Metabase keeps appearing

The Metabase entry is the second time the platform has featured in exploited-flaw reporting recently, and the reason is structural rather than a reflection of code quality. Business intelligence tools hold live credentials for several production databases at once, because that is what they need to draw a chart. A SQL injection there is not a route to the application, it is a route to the data.

Self-hosted analytics software also tends to lack an owner who watches security advisories, which is the same dynamic behind [the Metabase zero-day exploited earlier in August](/article/metabase-zero-day-august-2026/). The practical mitigation is to narrow what the tool can reach: a reporting account with read-only rights on the specific schemas reports actually use limits the damage from an injection flaw far more than patch speed alone does.

### Check whether you are exposed through someone else

One question worth asking that does not appear on any advisory: which of these products is being run on your behalf rather than by you? An MSP's N-central instance, a managed firewall, a hosted analytics tool. You cannot patch those, but you can ask the provider whether they have, and the answer belongs in your risk register either way.

## The short version for patching

If you run any of the first three and have not patched, that is today's job rather than this sprint's. CVE-2026-68820 should follow immediately after, and state-linked operators have already used that same WinSock flaw to [drop kernel rootkits on aerospace firms](/article/lazarus-windows-zero-day-cve-2026-68820/) once they had a foothold, which is what [privilege escalation to SYSTEM](/article/what-is-privilege-escalation-attack/) buys them.

Two additions beyond patching. For the three internet-facing items, confirm whether the management interface needs to be reachable from the internet at all, because for most deployments it does not and restricting it removes the attack path regardless of patch state. And for anything that was exposed while exploitation was underway, patch and then hunt, because a fix evicts nobody who is already resident.

The timing argument is straightforward. [Mandiant found](https://cloud.google.com/blog/topics/threat-intelligence/time-to-exploit-trends-2023) the average time-to-exploit fell to five days in 2023, from 32 days in 2021-22, an 84% reduction in grace period, and Verizon attributes roughly 20% of breaches to vulnerability exploitation. With IBM putting average breach costs near $4.88 million and Sophos reporting mean ransomware recovery around $2.73 million before any ransom, an emergency change window is comfortably the cheaper option.

## Key takeaways

- CVE-2026-18577, an authentication bypass in N-able N-central, reached CISA's exploited list after real customer compromises rather than a research write-up.
- KEV inclusion means proven exploitation, not high severity. It is the one feed where listing implies somebody has already been attacked this way.
- Cisco Secure Firewall (CVE-2026-20349) and a Metabase SQL injection (CVE-2026-72898) joined in the same week, with CVE-2026-68820 in Windows WinSock exploited for SYSTEM escalation.
- Triage by reachability, not score. The three internet-facing flaws give attackers entry; the WinSock bug only matters once they are inside.
- Around 75% of this batch is management or security tooling, which is the current shape of initial access: go through the thing that watches the door.
- RMM compromise multiplies. Past incidents reached roughly 1,500 businesses via about 60 MSPs, and a trojanised monitoring update hit some 18,000 organisations.
- Patch today, take management interfaces off the internet where possible, and hunt for persistence on anything exposed during active exploitation.

## Sources

- [The Hacker News, CISA adds exploited N-able N-central flaw to KEV catalog](https://thehackernews.com/2026/08/cisa-adds-exploited-n-able-n-central.html)
- [CISA, Three known exploited vulnerabilities added to catalog](https://www.cisa.gov/news-events/alerts/2026/08/11/cisa-adds-three-known-exploited-vulnerabilities-catalog)
- [SecurityWeek, August 2026 Patch Tuesday, Microsoft fixes 421 CVEs](https://www.securityweek.com/august-2026-patch-tuesday-microsoft-fixes-421-cves-one-exploited-zero-day/)
