Last updated 14 November 2024

Incorrect control of environment variables in PostgreSQL PL/Perl allows an unprivileged database user to change sensitive process environment variables (e.g. PATH). That often suffices to enable arbitrary code execution, even if the attacker lacks a database server operating system user. Versions before PostgreSQL 17.1, 16.5, 15.9, 14.14, 13.17, and 12.21 are affected.

A flaw was found in PostgreSQL versions before 13.1, before 12.5, before 11.10, before 10.15, before 9.6.20 and before 9.5.24. An attacker having permission to create non-temporary objects in at least one schema can execute arbitrary SQL functions under the identity of a superuser. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.

A flaw was found in postgresql in versions before 13.3, before 12.7, before 11.12, before 10.17 and before 9.6.22. While modifying certain SQL array values, missing bounds checks let authenticated database users write arbitrary bytes to a wide area of server memory. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.

A flaw was found in PostgreSQL. There is an issue with incomplete efforts to operate safely when a privileged user is maintaining another user's objects. The Autovacuum, REINDEX, CREATE INDEX, REFRESH MATERIALIZED VIEW, CLUSTER, and pgamcheck commands activated relevant protections too late or not at all during the process. This flaw allows an attacker with permission to create non-temporary objects in at least one schema to execute arbitrary SQL functions under a superuser identity.

An extension script is vulnerable if it uses @extowner@, @extschema@, or @extschema:...@ inside a quoting construct (dollar quoting, '', or ""). No bundled extension is vulnerable. Vulnerable uses do appear in a documentation example and in non-bundled extensions. Hence, the attack prerequisite is an administrator having installed files of a vulnerable, trusted, non-bundled extension. Subject to that prerequisite, this enables an attacker having database-level CREATE privilege to execute arbitrary code as the bootstrap superuser. PostgreSQL will block this attack in the core server, so there's no need to modify individual extensions.

Supported, Vulnerable Versions: 11 - 15.

A flaw was found in PostgreSQL that allows authenticated database users to execute arbitrary code through missing overflow checks during SQL array value modification. This issue exists due to an integer overflow during array modification where a remote user can trigger the overflow by providing specially crafted data. This enables the execution of arbitrary code on the target system, allowing users to write arbitrary bytes to memory and extensively read the server's memory.

A flaw was found in PostgreSQL versions before 13.1, before 12.5, before 11.10, before 10.15, before 9.6.20 and before 9.5.24. If a client application that creates additional database connections only reuses the basic connection parameters while dropping security-relevant parameters, an opportunity for a man-in-the-middle attack, or the ability to observe clear-text transmissions, could exist. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.

PostgreSQL is vulnerable to SQL injection. A remote attacker could send specially-crafted SQL statements when the server is configured to use trust authentication with a clientcert requirement or to use cert authentication, which could allow the attacker to view, add, modify or delete information in the back-end database.

Late privilege drop in REFRESH MATERIALIZED VIEW CONCURRENTLY in PostgreSQL allows an object creator to execute arbitrary SQL functions as the command issuer. The command intends to run SQL functions as the owner of the materialized view, enabling safe refresh of untrusted materialized views. The victim is a superuser or member of one of the attacker's roles. The attack requires luring the victim into running REFRESH MATERIALIZED VIEW CONCURRENTLY on the attacker's materialized view. As part of exploiting this vulnerability, the attacker creates functions that use CREATE RULE to convert the internally-built temporary table to a view. Versions before PostgreSQL 15.6, 14.11, 13.14, and 12.18 are affected. The only known exploit does not work in PostgreSQL 16 and later. For defense in depth, PostgreSQL 16.2 adds the protections that older branches are using to fix their vulnerability.

A vulnerability was found in PostgreSQL. This attack requires permission to create non-temporary objects in at least one schema, the ability to lure or wait for an administrator to create or update an affected extension in that schema, and the ability to lure or wait for a victim to use the object targeted in CREATE OR REPLACE or CREATE IF NOT EXISTS. Given all three prerequisites, this flaw allows an attacker to run arbitrary code as the victim role, which may be a superuser.

A flaw was found in the psql interactive terminal of PostgreSQL in versions before 13.1, before 12.5, before 11.10, before 10.15, before 9.6.20 and before 9.5.24. If an interactive psql session uses \gset when querying a compromised server, the attacker can execute arbitrary code as the operating system account running psql. The highest threat from this vulnerability is to data confidentiality and integrity as well as system availability.

A flaw was found in PostgreSQL, where some PostgreSQL extensions did not use the searchpath safely in their installation script. This flaw allows an attacker with sufficient privileges to trick an administrator into executing a specially crafted script during the extension's installation or update. The highest threat from this vulnerability is to confidentiality, integrity, as well as system availability.

The Windows installer for PostgreSQL 9.5 - 12 invokes system-provided executables that do not have fully-qualified paths. Executables in the directory where the installer loads or the current working directory take precedence over the intended executables. An attacker having permission to add files into one of those directories can use this to execute arbitrary code with the installer's administrative rights.

CVE-2023-2454: CREATE SCHEMA ... schemaelement defeats protective searchpath changes.

Versions Affected: 11 - 15. This problem is quite old.

This enabled an attacker having database-level CREATE privilege to execute arbitrary code as the bootstrap superuser. Database owners have that right by default, and explicit grants may extend it to other users.

https://www.postgresql.org/support/security/CVE-2023-2454/

A flaw was found in PostgreSQL, where it did not properly sanitize the searchpath during logical replication. This flaw allows an authenticated attacker to use this flaw in an attack similar to CVE-2018-1058 to execute an arbitrary SQL command in the user's context for replication. The highest threat from this vulnerability is to confidentiality, integrity, as well as system availability.

A flaw was found in PostgreSQL's "ALTER ... DEPENDS ON EXTENSION", where sub-commands did not perform authorization checks. An authenticated attacker could use this flaw in certain configurations to perform drop objects such as function, triggers, et al., leading to database corruption.

A flaw was found in postgresql. Using an INSERT ... ON CONFLICT ... DO UPDATE command on a purpose-crafted table, an authenticated database user could read arbitrary bytes of server memory. The highest threat from this vulnerability is to data confidentiality.

A flaw was found in postgresql. Using an UPDATE ... RETURNING command on a purpose-crafted table, an authenticated database user could read arbitrary bytes of server memory. The highest threat from this vulnerability is to data confidentiality.

A flaw was found in postgresql. A purpose-crafted query can read arbitrary bytes of server memory. In the default configuration, any authenticated database user can complete this attack at will. The attack does not require the ability to create objects. If server settings include maxworkerprocesses=0, the known versions of this attack are infeasible. However, undiscovered variants of the attack may be independent of that setting.

A man-in-the-middle attacker can inject false responses to the client's first few queries, despite the use of SSL certificate verification and encryption.

Incomplete tracking in PostgreSQL of tables with row security allows a reused query to view or change different rows from those intended. CVE-2023-2455 and CVE-2016-2193 fixed most interaction between row security and user ID changes. They missed cases where a subquery, WITH query, security invoker view, or SQL-language function references a table with a row-level security policy. This has the same consequences as the two earlier CVEs. That is to say, it leads to potentially incorrect policies being applied in cases where role-specific policies are used and a given query is planned under one role and then executed under other roles. This scenario can happen under security definer functions or when a common user and query is planned initially and then re-used across multiple SET ROLEs. Applying an incorrect policy may permit a user to complete otherwise-forbidden reads and modifications. This affects only databases that have used CREATE POLICY to define a row security policy. An attacker must tailor an attack to a particular application's pattern of query plan reuse, user ID changes, and role-specific row security policies. Versions before PostgreSQL 17.1, 16.5, 15.9, 14.14, 13.17, and 12.21 are affected.

CVE-2023-2455: Row security policies disregard user ID changes after inlining.

Versions Affected: 11 - 15. This problem is quite old.

While CVE-2016-2193 fixed most interaction between row security and user ID changes, it missed a scenario involving function inlining. This leads to potentially incorrect policies being applied in cases where role-specific policies are used and a given query is planned under one role and then executed under other roles. This scenario can happen under security definer functions or when a common user and query is planned initially and then re-used across multiple SET ROLEs. Applying an incorrect policy may permit a user to complete otherwise-forbidden reads and modifications. This affects only databases that have used CREATE POLICY to define a row security policy.

https://www.postgresql.org/support/security/CVE-2023-2455/

A flaw was found in PostgreSQL involving the pgcancelbackend role that signals background workers, including the logical replication launcher, autovacuum workers, and the autovacuum launcher. Successful exploitation requires a non-core extension with a less-resilient background worker and would affect that specific background worker only. This issue may allow a remote high privileged user to launch a denial of service (DoS) attack.

A user having some UPDATE privilege on a partitioned table but lacking the SELECT privilege on some column may nonetheless be able to acquire denied-column values from an error message. This is similar to CVE-2014-8161, but the conditions for exploitation are more rare.

The PostgreSQL project thanks Heikki Linnakangas for reporting this problem.

A memory disclosure vulnerability was found in PostgreSQL that allows remote users to access sensitive information by exploiting certain aggregate function calls with 'unknown'-type arguments. Handling 'unknown'-type values from string literals without type designation can disclose bytes, potentially revealing notable and confidential information. This issue exists due to excessive data output in aggregate function calls, enabling remote users to read some portion of system memory.

Incorrect privilege assignment in PostgreSQL allows a less-privileged application user to view or change different rows from those intended. An attack requires the application to use SET ROLE, SET SESSION AUTHORIZATION, or an equivalent feature. The problem arises when an application query uses parameters from the attacker or conveys query results to the attacker. If that query reacts to currentsetting('role') or the current user ID, it may modify or return data as though the session had not used SET ROLE or SET SESSION AUTHORIZATION. The attacker does not control which incorrect user ID applies. Query text from less-privileged sources is not a concern here, because SET ROLE and SET SESSION AUTHORIZATION are not sandboxes for unvetted queries. Versions before PostgreSQL 17.1, 16.5, 15.9, 14.14, 13.17, and 12.21 are affected.

Client use of server error message in PostgreSQL allows a server not trusted under current SSL or GSS settings to furnish arbitrary non-NUL bytes to the libpq application. For example, a man-in-the-middle attacker could send a long error message that a human or screen-scraper user of psql mistakes for valid query results. This is probably not a concern for clients where the user interface unambiguously indicates the boundary between one error message and other text. Versions before PostgreSQL 17.1, 16.5, 15.9, 14.14, 13.17, and 12.21 are affected.

A modified, unauthenticated server can send an unterminated string during the establishment of Kerberos transport encryption. When a libpq client application has a Kerberos credential cache and doesn't explicitly disable option "gssencmode", a server can cause libpq to over-read and report an error message containing uninitialized bytes from and following its receive buffer. If libpq's caller somehow makes that message accessible to the attacker, this achieves a disclosure of the over-read bytes. We have not confirmed or ruled out viability of attacks that arrange for a crash or for presence of notable, confidential information in disclosed bytes.