On this page (6 sections)
CVE-2026-116945 is a heap-based buffer overflow in SQL Server that, according to its NVD record, allows an unauthorized attacker to execute code over a network. The record was published on 2026-11-10 and carries a critical CVSS score of 9.8.
Anyone who runs SQL Server should pay attention, especially if an instance is reachable from the internet or from many internal networks. The public information is still very thin, so this article separates what is confirmed from what you will need to verify in Microsoft's own advisory.
What is confirmed so far
The only primary data available is the NVD entry, which gives a one-sentence description: a heap-based buffer overflow in SQL Server allows an unauthorized attacker to execute code over a network. The record lists Microsoft's update guide page for this CVE as its reference.
Only one independent publisher covers this story so far. That means there is no second source confirming details such as affected editions, patch identifiers or attacker activity. Treat anything beyond the NVD description as unconfirmed until Microsoft's advisory says otherwise.
- CVSS score in the NVD record
- 9.8
- Date the record was published
- 2026-11-10
What a heap-based buffer overflow means
A heap buffer overflow happens when a program writes more data into a block of memory than that block was sized to hold. Because the memory sits on the heap, the extra data can overwrite neighboring structures that the program later trusts. In the worst case, an attacker shapes that data so the program runs code of the attacker's choosing.
Here the description says the attacker is unauthorized, so no valid login is needed, and the code runs over a network. That combination is consistent with the critical 9.8 score. A database server often holds the most valuable data a business has, and code execution on it can mean data theft, tampering or a foothold for moving deeper into a network.
Who is realistically exposed
- Servers with SQL Server ports reachable from the internet, which is the highest-risk situation.
- Instances reachable from large internal networks, VPN pools or partner connections, where one compromised workstation could reach the database.
- Hosting providers and agencies that run SQL Server for several customer sites on shared infrastructure.
- Websites and applications backed by SQL Server, because a database compromise affects the site on top of it.
- Forgotten or unmanaged test and development instances that never receive updates.
Hosted database services managed by a cloud provider may be patched by that provider, but the research does not confirm this for this CVE. Check your provider's notices instead of assuming.
How to check your exposure
Start with an inventory. You cannot judge exposure until you know every SQL Server instance you run and which networks can reach it.
This read-only T-SQL query, run in a SQL Server client such as sqlcmd or SQL Server Management Studio against each instance, returns the full version string, edition and patch level so you can compare it against the fixed build in Microsoft's advisory:
SELECT @@VERSION AS full_version,
SERVERPROPERTY('ProductVersion') AS product_version,
SERVERPROPERTY('ProductLevel') AS product_level,
SERVERPROPERTY('Edition') AS edition;To see whether a Windows host is listening for SQL Server connections, run this in an elevated PowerShell session on the server. It only lists listening TCP sockets; look for port 1433 or any custom port your instance uses and note whether the listener address is a public or broad interface:
Get-NetTCPConnection -State Listen | Sort-Object LocalPort | Format-Table LocalAddress, LocalPort, OwningProcessThen confirm from outside your network whether the database port answers. Review your firewall and security group rules for any inbound allow rule to the SQL Server port, and, where your policy permits, test the public addresses you own from an external host with a network scanning tool you already trust. An open database port facing the internet is worth closing regardless of this CVE.
What to do now
- Open Microsoft's advisory for CVE-2026-116945 and note the affected versions and the fixed build or update for your edition.
- List every SQL Server instance you run, including test, development and forgotten ones, and record each version with the query above.
- Install the update Microsoft lists for your version as soon as you can test it, starting with instances that are reachable from the internet.
- Until patched, block internet access to SQL Server ports at the firewall and allow only the application servers and admin hosts that need to connect.
- Make sure web applications connect to the database over a private network, not a public address.
- Take a verified backup before patching, and confirm you can restore it.
- Review SQL Server and Windows logs for unexpected restarts, new logins, new accounts or unusual outbound connections from the database host.
- Re-run the version query after patching to confirm the new build is active.
Detection and monitoring afterwards
The research does not say whether this flaw is being exploited, and the record shows no CISA Known Exploited Vulnerabilities listing. That can change quickly for a bug with a 9.8 score, so recheck the KEV catalog and Microsoft's page in the coming days.
- Watch for crashes or unexpected restarts of the SQL Server service, which can accompany memory corruption attempts.
- Alert on new or changed administrator logins, new Windows accounts and new scheduled tasks on the database host.
- Monitor outbound connections from database servers. They rarely need to reach the internet, so new destinations deserve a look.
- Keep an eye on the websites that depend on the database: sudden errors, new admin users or changed content can be the first visible sign of trouble. Webforta Monitor can track uptime and certificate or DNS changes on those sites, which helps you notice side effects quickly.
If you suspect a database host was compromised, follow a structured response such as the one in Website hacked: what to do, and rotate credentials that the server stored or used.
Sources
This article is based on the public sources below. Details can change as vendors and researchers publish more; check the original advisories before acting on version numbers.
