PostgreSQL versions 10.x before 10.9 and versions 11.x before 11.4 are vulnerable to a stack-based buffer overflow. Any authenticated user can overflow a stack-based buffer by changing the user's own password to a purpose-crafted value. This often suffices to execute arbitrary code as the PostgreSQL operating system account.

PostgreSQL 9.2.x before 9.2.3, 9.1.x before 9.1.8, 9.0.x before 9.0.12, 8.4.x before 8.4.16, and 8.3.x before 8.3.23 does not properly declare the enum_recv function in backend/utils/adt/enum.c, which causes it to be invoked with incorrect arguments and allows remote authenticated users to cause a denial of service (server crash) or read sensitive process memory via a crafted SQL command, which triggers an array index error and an out-of-bounds read.

The PL/perl and PL/Tcl implementations in PostgreSQL 7.4 before 7.4.30, 8.0 before 8.0.26, 8.1 before 8.1.22, 8.2 before 8.2.18, 8.3 before 8.3.12, 8.4 before 8.4.5, and 9.0 before 9.0.1 do not properly protect script execution by a different SQL user identity within the same session, which allows remote authenticated users to gain privileges via crafted script code in a SECURITY DEFINER function, as demonstrated by (1) redefining standard functions or (2) redefining operators, a different vulnerability than CVE-2010-1168, CVE-2010-1169, CVE-2010-1170, and CVE-2010-1447.

The "make check" command for the test suites in PostgreSQL 9.3.3 and earlier does not properly invoke initdb to specify the authentication requirements for a database cluster to be used for the tests, which allows local users to gain privileges by leveraging access to this cluster.

PostgreSQL 8.1.x before 8.1.4, 8.0.x before 8.0.8, 7.4.x before 7.4.13, 7.3.x before 7.3.15, and earlier versions allows context-dependent attackers to bypass SQL injection protection methods in applications via invalid encodings of multibyte characters, aka one variant of "Encoding-Based SQL Injection."

Multiple buffer overflows in gram.y for PostgreSQL 8.0.1 and earlier may allow attackers to execute arbitrary code via (1) a large number of variables in a SQL statement being handled by the read_sql_construct function, (2) a large number of INTO variables in a SELECT statement being handled by the make_select_stmt function, (3) a large number of arbitrary variables in a SELECT statement being handled by the make_select_stmt function, and (4) a large number of INTO variables in a FETCH statement being handled by the make_fetch_stmt function, a different set of vulnerabilities than CVE-2005-0245.

Improper enforcement of message integrity in PostgreSQL GSSAPI support allows a user to negotiate GSSAPI contrary to pg_hba.conf rules, via initial direct TLS connection. Despite a pg_hba.conf that appears to require GSSAPI, the connection may exchange data over TLS encryption alone. If the TLS settings are more permissive than the GSS settings, the connection may continue with lesser protection. Within major versions 17-18, minor versions before PostgreSQL 18.5 and 17.11 are affected. Versions before PostgreSQL 17 are unaffected.

backend/tcop/postgres.c in PostgreSQL 8.1.x before 8.1.5 allows remote authenticated users to cause a denial of service (daemon crash) related to duration logging of V3-protocol Execute messages for (1) COMMIT and (2) ROLLBACK SQL statements.

Incomplete tracking in PostgreSQL of changes to role membership, role attributes, and database ownership allows a query to continue using cached row-level security policies after those changes require a different policy, via plan reuse. Stale policies continue until some other event invalidates the cache or connection termination ends the session. This permits a user to complete reads and modifications that were recently permitted but now forbidden. An attacker must tailor an attack to a particular application's pattern of privilege removal and role-specific row security policies. Versions before PostgreSQL 18.5, 17.11, 16.15, 15.19, and 14.24 are affected.

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.

The xml_parse function in the libxml2 support in the core server component in PostgreSQL 8.3 before 8.3.20, 8.4 before 8.4.13, 9.0 before 9.0.9, and 9.1 before 9.1.5 allows remote authenticated users to determine the existence of arbitrary files or URLs, and possibly obtain file or URL content that triggers a parsing error, via an XML value that refers to (1) a DTD or (2) an entity, related to an XML External Entity (aka XXE) issue.

PostgreSQL PL/Java after 9.0 does not honor access controls on large objects.

PostgreSQL before 8.3.7, 8.2.13, 8.1.17, 8.0.21, and 7.4.25 allows remote authenticated users to cause a denial of service (stack consumption and crash) by triggering a failure in the conversion of a localized error message to a client-specified encoding, as demonstrated using mismatched encoding conversion requests.

Buffer over-read in PostgreSQL function pg_restore_attribute_stats() accepts array values of unmatched length, which causes query planning to read past end of one array. This allows a table maintainer to infer memory values past that array end. Within major version 18, minor versions before PostgreSQL 18.4 are affected. Versions before PostgreSQL 18 are unaffected.

PostgreSQL 8.1.x before 8.1.4, 8.0.x before 8.0.8, 7.4.x before 7.4.13, 7.3.x before 7.3.15, and earlier versions allows context-dependent attackers to bypass SQL injection protection methods in applications that use multibyte encodings that allow the "\" (backslash) byte 0x5c to be the trailing byte of a multibyte character, such as SJIS, BIG5, GBK, GB18030, and UHC, which cannot be handled correctly by a client that does not understand multibyte encodings, aka a second variant of "Encoding-Based SQL Injection." NOTE: it could be argued that this is a class of issue related to interaction errors between the client and PostgreSQL, but a CVE has been assigned since PostgreSQL is treating this as a preventative measure against this class of problem.

Improper validation of type "oidvector" in PostgreSQL allows a database user to disclose a few bytes of server memory. We have not ruled out viability of attacks that arrange for presence of confidential information in disclosed bytes, but they seem unlikely. Versions before PostgreSQL 18.2, 17.8, 16.12, 15.16, and 14.21 are affected.

PostgreSQL 8.0.x before 8.0.6 and 8.1.x before 8.1.2, when running on Windows, allows remote attackers to cause a denial of service (postmaster exit and no new connections) via a large number of simultaneous connection requests.

The Safe (aka Safe.pm) module 2.26, and certain earlier versions, for Perl, as used in PostgreSQL 7.4 before 7.4.29, 8.0 before 8.0.25, 8.1 before 8.1.21, 8.2 before 8.2.17, 8.3 before 8.3.11, 8.4 before 8.4.4, and 9.0 Beta before 9.0 Beta 2, allows context-dependent attackers to bypass intended (1) Safe::reval and (2) Safe::rdo access restrictions, and inject and execute arbitrary code, via vectors involving subroutine references and delayed execution.

PostgreSQL versions 8.4 - 9.6 are vulnerable to information leak in pg_user_mappings view which discloses foreign server passwords to any user having USAGE privilege on the associated foreign server.

PostgreSQL 8.3.x before 8.3.19, 8.4.x before 8.4.12, 9.0.x before 9.0.8, and 9.1.x before 9.1.4 allows remote authenticated users to cause a denial of service (server crash) by adding the (1) SECURITY DEFINER or (2) SET attributes to a procedural language's call handler.