PostgreSQL’s 12 November Update Fixes 52 Security Flaws and Ends Support for Version 14
PostgreSQL released fixes for 52 security flaws on 12 November and made 14.25 the last release of version 14. Who can reach the bugs, what to reindex, and why to take a fresh full backup.

In this article
Table of Contents
Table of contents
On Thursday 12 November the PostgreSQLThe KBY LexiconPostgreSQLPostgreSQL is an open-source relational database management system queried with SQL. project released updates for every supported version of its database, 18.7, 17.12, 16.16 and 15.20 among them, and fixed 52 security vulnerabilities in the process. The same day it shipped 14.25, which the project’s versioning policy lists as the final release of PostgreSQL 14. Anyone still running 14 now has a database that will get no more security fixes, and 40 of the 52 flaws affect it. It is the kind of release that gets a line in a changelog feed and then waits in a queue, which is why we are covering it five days on.
The short version for IT teams: if you run PostgreSQL 15 to 18, install the new minor release on your normal change cycle, sooner where untrusted people can log in to the database. If you run 14, you now have a migration to schedule, not a patch to apply. Either way, read the release notes before you update, because this release also asks for some reindexing and a fresh full backupThe KBY LexiconBackupA causally disconnected, point-in-time copy of system state, tagged with a consistency marker, that lets you recover from logical corruption or data loss independent of the source system's health. .
#What was fixed, in numbers
The project’s security page lists 52 CVEs fixed in the 12 November releases. Counting them from that page, 32 carry a CVSS score of 8.8, the highest in the batch, and 38 score 7.0 or above. The lowest is 3.1. Help Net Security, which reported the release on Friday, gives the same 32 and the same low of 3.1.
Most of the serious ones are memory-safety or type-confusion bugs in the server, the kind where PostgreSQL treats a value as the wrong data type or uses memory after freeing it. The project’s own titles for them include “plpgsql record-field type confusion executes arbitrary code”, “expanded-record domain-check use-after-free executes arbitrary code” and “dblink reentrant disconnect use-after-free executes arbitrary code”. Forty-three of the 52 are in the core server, three are in contrib modules, and six are in client tools.
The release notes lead with a different change: a set of defences against altering a composite type while queries are using it (CVE-2026-6460). The notes say full protection means rebuilding C-language extensions, and that some extensions will need source-code changes. A new setting, composite_type_mismatch, can switch the checks off, but the notes warn that doing so carries security risks. If you run in-house or third-party C extensions, that is the item most likely to cost you time.
#Who can actually reach these bugs
Almost all of the 52 need a database account. The project’s scoring marks only five as needing no privileges at all. Three of those five need a person to do something first, such as run a restore or accept a malicious replication target. A fourth is a pg_basebackup format-string flaw scored 3.7, which the release notes say can only be reached if files with unusual names exist in the server’s data directory. The fifth is a pre-authentication cancel-request flaw scored 5.3: the release notes say an attacker could, with good timing, make the process that receives a cancel request exit while holding a spinlock, which crashes and restarts the database.
That shapes the risk. A PostgreSQL server that only your own application connects to, with one trusted role, is a harder target than the scores suggest. A server where many users or tenants can run SQL, or where developers and contractors have direct access, is the case these fixes matter for. We would also treat any database reachable from the internet as urgent, whatever the scores say, because the cancel-request flaw is scored as needing no login.
#The client tools are part of the patch
It is easy to patch the server and forget the machine that runs backups. The release fixes problems in pg_dump, pg_restore, reindexdb, vacuumdb and pg_basebackup. Help Net Security singles out a SQL injection in pg_dump through not-null constraint names, a tar-format restore script that failed to block psql command injection, and mishandled quoting of index names in reindexdb. Its advice is to update the client binaries along with the server if you run these tools against databases where other users can create and name objects.
In practice that means the backup host, the CI runners that restore test databases, and anyone’s laptop with a psql install. Run psql --version on each, not just SELECT version(); on the server.
#Reindexing and backups: the part that is easy to miss
The release notes’ migration section says a dump and restore is not required for 18.x users, but then adds two warnings. Several bugs could have caused index corruption, so you should read the changelog entries that mention REINDEX and reindex anything affected. And existing incremental backup files “should not be trusted”, because there are several ways they could have become corrupt. The project recommends taking a new full backup after updating and basing later incrementals on it. The 14.25 notes carry the same warning about reindexing.
The affected cases the notes and Help Net Security name are specific. B-tree indexes on columns declared as bpchar with no length need a reindex, because index de-duplication is unsafe for that type; the ordinary CHAR(n) is not at risk. A bug in a TOAST-table index could cause “missing chunk number” errors, and the notes say a REINDEX will recover from that, since only the index is damaged. Help Net adds GIN indexes suspected of corruption and pg_trgm GiST indexes that use gist_trgm_ops.
Our advice is to take the new full backup first, as soon as the binaries are in place, because it costs a night of storage and it is the step people skip. Reindexing can be scheduled for a quiet window, REINDEX CONCURRENTLY where you cannot take locks.
#PostgreSQL 14 has reached the end
The versioning policy says each major version is supported for five years after its first release, after which a final minor version is released and the software is unsupported. The table on the project’s page shows 14 first released on 30 September 2021 with its final release on 12 November 2026. Version 13 and everything older is already marked unsupported. The dates for the others are 11 November 2027 for 15, 9 November 2028 for 16, 8 November 2029 for 17 and 14 November 2030 for 18.
This matters more than a usual end-of-life notice. The final 14.x release fixes 40 of these 52 CVEs. The next time the project finds a flaw like these, version 14 will not get the fix, and the people who find flaws in PostgreSQL are clearly finding a lot of them: the release notes credit, among many individual researchers, the OpenAI Security Research Team, OpenAI Codex Security, the Anthropic OSS programme, and Claude with Ada Logics. We read that as a sign that AI-assisted code review is raising the number of bugs found in mature open-source projects, and so the number of fixes that unsupported versions will miss. That is our reading, not something the project states.
Managed services have their own calendars, and we have not checked them here. If your database runs on a cloud provider’s PostgreSQL service, look at that provider’s own version-support page rather than assuming it follows the community’s date.
#What we would do this week
- Find every PostgreSQL instance and client install you own and record the version. Start with anything that holds data you cannot lose, then anything reachable from outside.
- For 15 to 18, schedule the minor update (a stop, install and restart; the policy says minor upgrades need no dump and restore). Put databases where many people can run SQL first.
- After updating, take a new full backup and start the incremental chain from it. Treat older incrementals as untrusted for restore planning.
- Work through the changelog lines that mention
REINDEXand decide which indexes you have that match. - If you run C extensions, test them against the new release before production, and do not turn off
composite_type_mismatchto make a failing test pass. - For every 14 instance, set a migration date. Moving between major versions needs
pg_upgradeor a dump and reload, so it is a project, and it is easier to plan before the next 14 flaw is published than after.
One date to keep in view: Help Net Security reports that the project shipped PostgreSQL 19 RC 2 alongside this release, with general availability planned for 19 November. It also notes that RC 2 reverts support for JSON_ARRAY() and JSON_OBJECT() over arrays, composite types and range types in expression indexes, generated columns and partition keys, so objects built that way need to be dropped or redefined before upgrading. If you were testing 19 for a 14 migration, check for that. We would not move a production system to a brand-new major version in its first week, and we would rather land on 17 or 18 than on 15, which has about a year left.
Evidence trail
Sources and verification
Primary documentation and external technical references used in this article.
About this page
Credited to
Dates
Published
Corrections
No corrections are listed for this page. Some review sweeps corrected pages without naming each one; those are in the corrections log.
Read-only scripts
- Windows devices with no BitLocker recovery key in Entra IDMicrosoft Entra ID, Windows
- Intune devices that haven't synced in N daysMicrosoft Intune
- Mailboxes forwarding to external addressesExchange Online
- BitLocker status of every volume on this PCWindows
Upcoming deadlines
- Intune moves to macOS 15 and laterQ4 2026
- SAP SuccessFactors provisioning: move from basic auth to workload identityNovember 2026
- Intune November release: JIT registration reminder at APNs renewal and automatic JIT for eligible tenantsNovember 2026
Explore KBY
The KBY Brief
A short email about twice a week with new runbooks, configuration traps and fixes from KBY Technologies.
SubscribeMore news
All newsDiscover more
Lexicon Definitions
Was this useful?
Reports go to the site owner and are used to correct pages. Corrections are listed at /corrections.
Get new fixes by email.
New runbooks, configuration traps and fixes from KBY Technologies, about twice a week.




Comments
Add a thoughtful note on PostgreSQL’s 12 November Update Fixes 52 Security Flaws and Ends Support for Version 14. Comments are checked for spam and held for moderation before appearing.