Veille CVE

Vendredi 10 juillet 2026

4 CVE

Critique 1 Élevée 2 Moyenne 0 Faible 1 Non évaluée 0 Exploitées (KEV) Tri : plus récentes « Postgresql postgresql » ✕
  1. CVE-2026-53448 7.2 Élevée

    coturn · Coturn project coturn

    Coturn is a free open source implementation of TURN and STUN Server. Prior to 4.12.0, the coturn HTTPS admin panel passes HTTP query parameters directly into SQL queries via snprintf string interpolation without sanitization. The is_secure_string filter that protects the STUN protocol path is not applied to the admin panel's delete-user, delete-secret, and delete-IP operations, so an authenticated admin can inject arbitrary SQL through the du, ds, and dip parameters, gaining full database control and potentially OS-level access via PostgreSQL COPY TO PROGRAM. This issue is fixed in version 4.12.0.

  2. CVE-2026-58225 2.1 Faible

    elixir-ecto postgrex

    SQL Injection vulnerability in elixir-ecto postgrex allows an attacker who can influence a LISTEN channel name to inject SQL into the reconnect replay query, causing a denial of service of the notification connection. Postgrex.Notifications sanitizes channel names with quote_channel/1, which doubles double quotes so the name is safe inside a double-quoted identifier. This protects the single-statement LISTEN and UNLISTEN paths. On every (re)connect, however, handle_connect/1 replays all registered channels at once by concatenating their LISTEN statements and wrapping them in a dollar-quoted anonymous code block (DO $$BEGIN ... END$$). quote_channel/1 does not escape the $$ dollar-quote delimiter that opens and closes this block. The listen/3 guards only reject null bytes and names longer than 63 bytes, so a channel name containing $$ passes validation unchanged. Once such a name is embedded, its $$ prematurely terminates the outer dollar-quoted string and PostgreSQL parses the remainder as additional top-level statements. Because handle_connect/1 runs on every (re)connect, the malformed replay query is rejected each time and the notification connection never re-establishes its subscriptions, silently dropping notifications for every channel sharing that connection. An application is affected when it passes untrusted input (for example a tenant or user identifier) as a channel name to Postgrex.Notifications.listen/3. The double-quote doubling prevents forming a fully valid injected statement, so arbitrary SQL execution is not possible, but the corrupted query reliably breaks the shared notification connection for all tenants, resulting in denial of service. This issue affects postgrex: from 0.16.0 before 0.22.3.

  3. CVE-2026-54760 9.3 Critique

    langroid

    Langroid is a framework for building large-language-model-powered applications. Prior to version 0.65.1, the `SQLChatAgent` SQL-injection mitigation, with default `allow_dangerous_operations=False`, combines a raw-text regex blocklist (`_DANGEROUS_SQL_PATTERNS`) with a `sqlglot` SELECT-only statement allowlist. The blocklist entries that target callable functions require the function name to be immediately followed by `\s*\(`. PostgreSQL accepts the same call with the name separated from `(` by a quoted identifier, an inline comment, or schema qualification. These forms evade the regex, still parse as `SELECT`, and execute the same PostgreSQL function. This restores the `pg_read_file` server-side file-read primitive that the prior CVE-2026-25879 / GHSA-pmch-g965-grmr fix was meant to block: the parent advisory fixed a missing `pg_read_file` blocklist entry, while this report shows that the added regex is bypassable. Version 0.65.1 fixes the issue.

  4. CVE-2026-50180 8.7 Élevée

    langroid

    Langroid is a framework for building large-language-model-powered applications. Prior to version 0.64.0, `SQLChatAgent` in `langroid` ships a `_validate_query` defense-in-depth layer whose `_DANGEROUS_SQL_PATTERNS` regex blocklist enumerates dangerous SQL primitives by specific function name. The list misses the canonical PostgreSQL filesystem-disclosure family `pg_read_file()`, `pg_stat_file()`, `pg_ls_logdir()`, `pg_ls_waldir()`, `pg_current_logfile()` (and similar `SELECT`-shaped functions in the same family). It also leaves SQL Server `OPENDATASOURCE` and SQLite `ATTACH '<file>' AS x` (DATABASE keyword omitted) unblocked. An attacker able to shape the LLM's generated SQL (directly via prompt input or transitively via prompt-injection in data the LLM ingests) can read arbitrary files from the PostgreSQL host through ordinary `SELECT` queries, even with the agent's strict default configuration (`allow_dangerous_operations=False`, `allowed_statement_types=['SELECT']`). The payloads survive the statement-type allowlist (each is a `SELECT`) and pass through the regex blocklist (none of the function names match), then reach the live SQLAlchemy engine via `SQLChatAgent.run_query`. Version 0.64.0 contains a patch for the issue.