AVideo has an OS Command Injection via Unescaped URL in LinkedIn Video Upload Shell Command

Description

Summary

The uploadVideoToLinkedIn() method in the SocialMediaPublisher plugin constructs a shell command by directly interpolating an upload URL received from LinkedIn's API response, without sanitization via escapeshellarg(). If an attacker can influence the LinkedIn API response (via MITM, compromised OAuth token, or API compromise), they can inject arbitrary OS commands that execute as the web server user.

Details

The vulnerability exists in plugin/SocialMediaPublisher/Objects/SocialUploader.php.

The initializeLinkedInUploadSession() method (line 649) sends a POST request to https://api.linkedin.com/rest/videos?action=initializeUpload and parses the JSON response at line 693:

// SocialUploader.php:693
$responseArray = json_decode($response, true);

The parsed uploadInstructions array is iterated at line 532, and each uploadUrl is passed to uploadVideoToLinkedIn() at line 542:

// SocialUploader.php:542
$uploadResponse = self::uploadVideoToLinkedIn($instruction['uploadUrl'], $tmpFile);

The uploadVideoToLinkedIn() method (line 711) constructs a shell command by directly concatenating both $uploadUrl and $filePath into a string passed to exec():

// SocialUploader.php:713-720
$shellCmd = 'curl -v -H "Content-Type:application/octet-stream" --upload-file "' .
    $filePath . '" "' .
    $uploadUrl . '" 2>&1';

_error_log("Upload Video Shell Command:\n" . $shellCmd);

exec($shellCmd, $o);

Neither $uploadUrl nor $filePath is sanitized with escapeshellarg(). A malicious URL such as https://uploads.linkedin.local" ; id ; echo " would break out of the quoted string and execute arbitrary commands.

The $uploadUrl originates from LinkedIn's API response — a trusted third-party source over HTTPS — so exploitation requires compromising that response (MITM at CA level, compromised OAuth token leading to attacker-controlled API responses, or LinkedIn API compromise). This makes the attack complexity high, but the missing sanitization is a defense-in-depth failure that could become critical if the trust boundary is ever violated.

PoC

This vulnerability requires manipulating the LinkedIn API response. A simulated proof-of-concept using a local proxy:

Step 1: Set up a proxy that intercepts the LinkedIn API response and replaces the uploadUrl field:

{
  "value": {
    "uploadInstructions": [
      {
        "uploadUrl": "https://example.com\" ; id > /tmp/pwned ; echo \"",
        "firstByte": 0,
        "lastByte": 1024
      }
    ],
    "uploadToken": "token123",
    "video": "urn:li:video:123"
  }
}

Step 2: The resulting shell command becomes:

curl -v -H "Content-Type:application/octet-stream" --upload-file "/tmp/tmpfile" "https://uploads.linkedin.local" ; id > /tmp/pwned ; echo "" 2>&1

Step 3: The id command executes as the web server user, writing output to /tmp/pwned.

Step 4: Verify:

cat /tmp/pwned
# uid=33(www-data) gid=33(www-data) groups=33(www-data)

Impact

  • Remote Code Execution: If the LinkedIn API response is compromised, an attacker gains arbitrary command execution as the web server user (www-data).
  • Confidentiality: Full read access to application source code, configuration files (including database credentials), and any data accessible to the web server process.
  • Integrity: Ability to modify application files, inject backdoors, or alter database records.
  • Practical risk is low due to the high attack complexity — exploitation requires compromising a trusted HTTPS API response from LinkedIn. This is primarily a defense-in-depth issue.

Recommended Fix

Sanitize both $uploadUrl and $filePath with escapeshellarg() before interpolation into the shell command. Alternatively, replace the exec() call with PHP's native cURL functions (which are already used elsewhere in the same class):

Option 1 — Minimal fix with escapeshellarg():

// plugin/SocialMediaPublisher/Objects/SocialUploader.php:711-715
static function uploadVideoToLinkedIn($uploadUrl, $filePath)
{
    $shellCmd = 'curl -v -H "Content-Type:application/octet-stream" --upload-file ' .
        escapeshellarg($filePath) . ' ' .
        escapeshellarg($uploadUrl) . ' 2>&1';

Option 2 — Replace shell exec with native PHP cURL (preferred):

static function uploadVideoToLinkedIn($uploadUrl, $filePath)
{
    $ch = curl_init();
    curl_setopt($ch, CURLOPT_URL, $uploadUrl);
    curl_setopt($ch, CURLOPT_HTTPHEADER, ['Content-Type: application/octet-stream']);
    curl_setopt($ch, CURLOPT_PUT, true);
    curl_setopt($ch, CURLOPT_INFILE, fopen($filePath, 'r'));
    curl_setopt($ch, CURLOPT_INFILESIZE, filesize($filePath));
    curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
    curl_setopt($ch, CURLOPT_HEADER, true);
    curl_setopt($ch, CURLOPT_VERBOSE, true);

    $response = curl_exec($ch);
    $headerSize = curl_getinfo($ch, CURLINFO_HEADER_SIZE);
    $headers = substr($response, 0, $headerSize);
    curl_close($ch);

    // Extract ETag from response headers
    $matches = [];
    preg_match('/(etag:)(\s?)(.*)(\n)/i', $headers, $matches);
    $etag = isset($matches[3]) ? trim($matches[3]) : null;

    // ... rest of function
}

Option 2 is strongly preferred as it eliminates the shell execution entirely, removing the injection surface and aligning with the PHP cURL usage already present in initializeLinkedInUploadSession() on line 664.

Basic information

Type
reviewed
Severity
medium
Advisory on GitHub
Open advisory ↗
Repository advisory
Open repository advisory ↗
Source code
Browse source ↗
Published (advisory)
2026-03-19 12:45:38 UTC
Updated
2026-03-25 18:33:54 UTC
GitHub reviewed
2026-03-19 12:45:38 UTC
NVD published
2026-03-22 17:17:09 UTC

EPSS Score

Score Percentile
0.03% 8.45%

CVSS Scores

Base score Version Severity Vector
5.9 3.1
CVSS:3.1/AV:N/AC:H/PR:H/UI:N/S:U/C:H/I:H/A:N Click to expand
Attack vector (AV:N)
Could be attacked over the internet or any normal routed network—not just someone sitting at the machine.
Attack complexity (AC:H)
Even with access, the exploit needs extra luck, timing, or a fussy environment to actually work.
Privileges required (PR:H)
They need powerful rights—admin, root, or similar—before this pays off.
User interaction (UI:N)
Nobody has to click “OK” or open a trap file; it can work without a victim helping.
Scope (S:U)
Damage stays in the same “trust bubble” as the broken component—no big spill into unrelated systems.
Confidentiality (C:H)
Serious risk that confidential data gets exposed in a big way.
Integrity (I:H)
They could widely tamper with or forge data—trust in the data is badly hurt.
Availability (A:N)
Service keeps running; no real outage angle.

Identifiers

CWEs

CWE id Name
CWE-78 Improper Neutralization of Special Elements used in an OS Command ('OS Command Injection')

Credits

  • offset (reporter)

Affected packages (1)

Vulnerable version ranges and first patched releases as published by GitHub.

Ecosystem Package Vulnerable range First patched Vulnerable functions
composer wwbn/avideo <= 25.0

References

cvelogic Threat Intelligence