Vulnerability Database

328,925

Total vulnerabilities in the database

CVE-2026-32813

Summary

The MyList configuration feature in Admidio allows authenticated users to define custom list column layouts. User-supplied column names, sort directions, and filter conditions are stored in the adm_list_columns table via prepared statements (safe storage), but are later read back and interpolated directly into dynamically constructed SQL queries without sanitization or parameterization. This is a classic second-order SQL injection: safe write, unsafe read.

An attacker can inject arbitrary SQL through these stored values to read, modify, or delete any data in the database, potentially achieving full database compromise.

Details

Step 1: Storing the Payload (Safe Write)

In modules/groups-roles/mylist_function.php (lines 89-115), user-supplied POST array values for column names, sort directions, and filter conditions are accepted. The only validation on column values is a prefix check (must start with usr_ or mem_). Sort and condition values have no validation at all. These values are stored in the database via ListConfiguration::addColumn() which calls Entity::save() using prepared statements -- so the INSERT/UPDATE is safe.

Key source file references:

  • D:\bugcrowd\admidio\repo\modules\groups-roles\mylist_function.php lines 89-115
  • D:\bugcrowd\admidio\repo\src\Roles\Entity\ListConfiguration.php lines 106-116

Step 2: Triggering the Payload (Unsafe Read)

When the list is viewed (via lists_show.php), ListConfiguration::getSql() reads the stored values and interpolates them directly into SQL in four locations:

Injection Point 1 -- lsc_special_field in SELECT clause: File D:\bugcrowd\admidio\repo\src\Roles\Entity\ListConfiguration.php lines 739-770. The lsc_special_field value is read from the database and used as a column name in the SELECT clause. Only three values (mem_duration, mem_begin, mem_end) get special handling; all others fall through to the default case where the raw value is used directly as both $dbColumnName and $sqlColumnName, then interpolated into the SQL as $dbColumnName AS $sqlColumnName.

Injection Point 2 -- lsc_sort in ORDER BY clause: File D:\bugcrowd\admidio\repo\src\Roles\Entity\ListConfiguration.php lines 790-792. The lsc_sort value is appended directly after the column name in the ORDER BY clause.

Injection Point 3 -- lsc_special_field in search conditions: File D:\bugcrowd\admidio\repo\src\Roles\Entity\ListConfiguration.php lines 611-621. The lsc_special_field value is interpolated into COALESCE() expressions used in search WHERE conditions.

Injection Point 4 -- lsc_filter via ConditionParser: File D:\bugcrowd\admidio\repo\src\Roles\ValueObject\ConditionParser.php line 347. The ConditionParser appends raw characters from the stored filter value to the SQL string. A single quote can break out of the SQL string context.

Root Cause

The addColumn() method and mylist_function.php accept arbitrary strings for column names, sort directions, and filter conditions. The only gate for column names is a prefix check (usr_ or mem_), which is trivially satisfied by an attacker (e.g., usr_id) UNION SELECT ...). No allowlist of valid column names exists. No server-side validation of sort values exists (should only allow ASC/DESC/empty). The frontend <select> element only offers ASC/DESC, but this is trivially bypassed by POSTing arbitrary values.

PoC

Prerequisites: Logged-in user with list edit permission (default: all logged-in users).

Step 1: Save a list config with SQL injection in lsc_special_field

curl -X POST "https://TARGET/adm_program/modules/groups-roles/mylist_function.php?mode=save_temporary" \ -H "Cookie: ADMIDIO_SESSION_ID=<session>" \ -d "adm_csrf_token=<csrf_token>" \ -d "column[]=usr_login_name" \ -d "column[]=usr_id FROM adm_users)--" \ -d "sort[]=" \ -d "sort[]=" \ -d "condition[]=" \ -d "condition[]=" \ -d "sel_roles[]=<valid_role_uuid>"

The second column value usr_id FROM adm_users)-- starts with usr_ so it passes the prefix check. When read back in getSql(), it is interpolated directly as a column expression in the SQL SELECT clause.

Step 2: Sort-based injection (simpler, no prefix check needed)

curl -X POST "https://TARGET/adm_program/modules/groups-roles/mylist_function.php?mode=save_temporary" \ -H "Cookie: ADMIDIO_SESSION_ID=<session>" \ -d "adm_csrf_token=<csrf_token>" \ -d "column[]=usr_login_name" \ -d "sort[]=ASC,(SELECT+CASE+WHEN+(1=1)+THEN+1+ELSE+1/0+END)" \ -d "condition[]=" \ -d "sel_roles[]=<valid_role_uuid>"

This injects into the ORDER BY clause. The sort value has zero server-side validation.

Step 3: The save_temporary mode automatically redirects to lists_show.php which calls ListConfiguration::getSql(), executing the injected SQL.

Impact

  • Data Exfiltration: An attacker can extract any data from the database including password hashes, email addresses, personal data of all members, and application configuration.
  • Data Modification: With stacked queries (supported by MySQL with PDO), the attacker can modify or delete data.
  • Privilege Escalation: Password hashes can be extracted and cracked, or admin accounts can be directly modified.
  • Full Database Compromise: The entire database is accessible through this vulnerability.

The attack requires authentication and CSRF token, but:

  1. Any logged-in user has this permission by default (when groups_roles_edit_lists = 1).
  2. The CSRF token is available in the same session.
  3. The injected payload persists in the database and triggers every time anyone views the list.

Fix 1: Allowlist for lsc_special_field

Add a strict allowlist of valid special field names before calling addColumn() in mylist_function.php. The list should match exactly the field names supported in getSql() and the JavaScript on mylist.php.

Fix 2: Validate lsc_sort values

In ListConfiguration::addColumn(), validate that the sort parameter is one of ASC, DESC, or empty string before storing it.

Fix 3: Defense-in-depth validation in ListConfiguration::getSql()

Also validate the lsc_special_field value against an allowlist in getSql() before interpolating it into the SQL string. This protects against payloads already stored in the database.

Fix 4: Escape filter values in ConditionParser

Use parameterized queries or at minimum escape single quotes in ConditionParser::makeSqlStatement().

CVSS v3:

  • Severity: Unknown
  • Score:
  • AV:N/AC:L/PR:L/UI:R/S:U/C:H/I:H/A:H

CWEs:

OWASP TOP 10:

Frequently Asked Questions

A security vulnerability is a weakness in software, hardware, or configuration that can be exploited to compromise confidentiality, integrity, or availability. Many vulnerabilities are tracked as CVEs (Common Vulnerabilities and Exposures), which provide a standardized identifier so teams can coordinate patching, mitigation, and risk assessment across tools and vendors.

CVSS (Common Vulnerability Scoring System) estimates technical severity, but it doesn't automatically equal business risk. Prioritize using context like internet exposure, affected asset criticality, known exploitation (proof-of-concept or in-the-wild), and whether compensating controls exist. A "Medium" CVSS on an exposed, production system can be more urgent than a "Critical" on an isolated, non-production host.

A vulnerability is the underlying weakness. An exploit is the method or code used to take advantage of it. A zero-day is a vulnerability that is unknown to the vendor or has no publicly available fix when attackers begin using it. In practice, risk increases sharply when exploitation becomes reliable or widespread.

Recurring findings usually come from incomplete Asset Discovery, inconsistent patch management, inherited images, and configuration drift. In modern environments, you also need to watch the software supply chain: dependencies, containers, build pipelines, and third-party services can reintroduce the same weakness even after you patch a single host. Unknown or unmanaged assets (often called Shadow IT) are a common reason the same issues resurface.

Use a simple, repeatable triage model: focus first on externally exposed assets, high-value systems (identity, VPN, email, production), vulnerabilities with known exploits, and issues that enable remote code execution or privilege escalation. Then enforce patch SLAs and track progress using consistent metrics so remediation is steady, not reactive.

SynScan combines attack surface monitoring and continuous security auditing to keep your inventory current, flag high-impact vulnerabilities early, and help you turn raw findings into a practical remediation plan.