This is not a typical SQL injection or privilege escalation vulnerability. The ctid selectivity estimator bug exploits a gap between PostgreSQL's type system and its internal memory layout: the function accepts generic input but processes it using ctid's internal memory representation, exposing raw 4-byte memory spans to any user who can create objects in the database. The practical impact depends on what those 4 bytes are part of—and that matters more than the precision loss itself. PostgreSQL doesn't scatter memory randomly; the planner heap contains tuple visibility markers, catalog entries, session parameters, and compiled query structures. A 4-byte read from planner state isn't reading noise—it's reading structured metadata that can serve as anchors for further exploitation, potentially including pointers, catalog references, or session tokens. The advisory's 'precision loss' phrasing obscures whether this is lossy floating-point conversion or partial struct exposure, but either way, the bounded and predictable nature of PostgreSQL's memory topology undermines any assumption that the precision limitation blocks practical exploitation. The 'object creator' prerequisite sounds narrow, but in modern deployments it's not: application service accounts routinely have schema creation rights, ORMs generate objects dynamically, migration scripts run under elevated contexts, and CI/CD pipelines hold object creation privileges. A compromised application account, a malicious developer with staging access, or a supply-chain-compromised deployment tool all satisfy this prerequisite. The blast radius extends to any deployment where the application and database share a trust relationship—which is the architectural default. Treat this as a planner-state information leak that enables reconnaissance for subsequent attacks, not as a constrained data exposure. Prioritize patching across all affected branches, and audit which roles in your deployment hold object creation privileges—these represent your actual attack surface.
CVE-2026-14668
Official description Straight from the sourceThe vendor's or NVD's own wording, published unedited. Authoritative, but often terse — it says what broke, rarely what to do.
NVD · uneditedType confusion regarding input of PostgreSQL ctid data type selectivity estimator allows an object creator to view a calculation derived from the value of an arbitrary 4-byte span of memory, via a chosen non-ctid input. While the calculation loses precision, substantial memory value recovery appears possible. Versions before PostgreSQL 18.5, 17.11, 16.15, 15.19, and 14.24 are affected.
Technical summary Written by usOur analysis, written from the advisory, the CVSS vector and the affected-version data. It adds context the advisory leaves out, and never invents facts that are not in the source.
dbcve analysisA resource is accessed as one type when it was actually allocated as another, so the code misreads memory layout — in interpreters and language runtimes this is frequently a direct path to code execution. It often arises from unchecked casts on attacker-influenced objects. The fix is strict type checks before casts and memory-safe access patterns.
General guidance for the type confusion class — the official description and references above are authoritative for this specific CVE. Want a bespoke review and a reviewed fix? Ask our team →
CVSS breakdown How the score is builtThe industry scoring standard. It rates how the flaw is reached, what it takes to exploit, and what an attacker gains — the score is derived from those, not the other way round.
From the vector- Attack vector
- Network
- Complexity
- Low
- Privileges
- Low
- User interaction
- None
- Scope
- Unchanged
- Confidentiality
- High
- Integrity
- None
- Availability
- High
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:H
Remediation Closing itWhat it takes to close this. Where a vendor fix exists we point at it; where none exists we say so plainly, and can build one. Effort estimates are scoped from the advisory, not from your codebase.
dbcve · scopedPostgreSQL 18.5, 17.11, 16.15, 15.19, or 14.24 (depending on your current major version branch)
- 1. Identify the current PostgreSQL version by running: SELECT version();
- 2. Determine which major version branch you're running (14, 15, 16, 17, or 18)
- 3. Plan maintenance window for the upgrade (PostgreSQL minor version upgrades are generally safe but require restart)
- 4. For minor version upgrades within the same major version (e.g., 17.x to 17.11): Download and install PostgreSQL 17.11 from the official PostgreSQL website or your package manager
- 5. For major version upgrades (e.g., 16.x to 17.x): Follow the PostgreSQL major version upgrade documentation, typically using pg_dumpall or pg_upgrade
- 6. After upgrading, restart the PostgreSQL service to load the new version
- 7. Verify the upgrade by running: SELECT version(); to confirm the new version number
Generated from the published advisory — verify against the referenced sources before acting.
There is no version to upgrade to and no patch to apply. Every affected install stays exposed until the vendor ships a fix — or somebody else builds one.
Free. We build fixes in the order the community asks for them — and we’ll tell you the moment this one lands.
We develop and verify an original fix where the vendor hasn’t, from $4,900. Deployed to your staging first — never straight to production.
Scope it with usSee what else the community needs solved on the solutions-needed board.
Scan for this in your stack
Free · runs locallyCheck whether your project pulls in CVE-2026-14668 — or any other known-vulnerable package — straight from your lock files. Free and open source; it runs locally and uploads nothing.
References Go to the primary sourcePrimary sources — vendor advisories, patches and trackers. Where our summary and a reference disagree, the reference wins.
Primary sourcesThis is not a typical SQL injection or privilege escalation vulnerability. The ctid selectivity estimator bug exploits a gap between PostgreSQL's type system and its internal memory layout: the function accepts generic input but processes it using ctid's internal memory representation, exposing raw 4-byte memory spans to any user who can create objects in the database. The practical impact depends on what those 4 bytes are part of—and that matters more than the precision loss itself. PostgreSQL doesn't scatter memory randomly; the planner heap contains tuple visibility markers, catalog entries, session parameters, and compiled query structures. A 4-byte read from planner state isn't reading noise—it's reading structured metadata that can serve as anchors for further exploitation, potentially including pointers, catalog references, or session tokens. The advisory's 'precision loss' phrasing obscures whether this is lossy floating-point conversion or partial struct exposure, but either way, the bounded and predictable nature of PostgreSQL's memory topology undermines any assumption that the precision limitation blocks practical exploitation. The 'object creator' prerequisite sounds narrow, but in modern deployments it's not: application service accounts routinely have schema creation rights, ORMs generate objects dynamically, migration scripts run under elevated contexts, and CI/CD pipelines hold object creation privileges. A compromised application account, a malicious developer with staging access, or a supply-chain-compromised deployment tool all satisfy this prerequisite. The blast radius extends to any deployment where the application and database share a trust relationship—which is the architectural default. Treat this as a planner-state information leak that enables reconnaissance for subsequent attacks, not as a constrained data exposure. Prioritize patching across all affected branches, and audit which roles in your deployment hold object creation privileges—these represent your actual attack surface.
Practitioner notes
ContributedPeer-ranked notes from engineers who’ve handled CVE-2026-14668 in production — separate from our analysis above.
The advisory tells you what broke. It rarely tells you what actually worked. If you’ve dealt with this one, that detail is what the next engineer is searching for.
- The version that genuinely resolved it — not the one the vendor claimed
- A config change or rule that shut the vector down
- A gotcha in the upgrade path that cost you an afternoon
A place for practitioners to share what actually worked: a mitigation you’ve tested, a configuration change, a version- or environment-specific caveat, or a link to a verified patch. The most useful notes rise to the top as peers upvote them, so the signal stays high.
- Verified mitigations, workarounds, and config changes
- Version or environment caveats, and links to real fixes
- No weaponised exploit code, or anything meant to cause harm
- No spam, self-promotion, credentials, or personal data