CVE-2024-48917: XXE in PHPSpreadsheet's XLSX reader

The XmlScanner class has a scan method which should prevent XXE attacks.

However, we found another bypass than the previously reported CVE-2024-47873, the regexes from the findCharSet method, which is used for determining the current encoding can be bypassed by using a payload in the encoding UTF-7, and adding at end of the file a comment with the value encoding="UTF-8" with ", which is matched by the first regex, so that encoding='UTF-7' with single quotes ' in the XML header is not matched by the second regex:

$patterns = [
'/encoding\\s*=\\s*"([^"]*]?)"/',
"/encoding\\s*=\\s*'([^']*?)'/",
];

A payload for the workbook.xml file can for example be created with CyberChef. If you open an Excel file containing the payload from the link above stored in the workbook.xml file with PhpSpreadsheet, you will receive an HTTP request on 127.0.0.1:12345. You can test that an HTTP request is created by running the nc -nlvp 12345 command before opening the file containing the payload with PhpSpreadsheet.

To create the payload you need:

1. Create a file containing <?xml version = "1.0" encoding='UTF-7' in an XML file
2. Use the link attached above to create your XXE payload and add it to the XML file.
3. Add +ADw-+ACE---encoding="UTF-8"--+AD4- to the end of the XML file, which is matched by the first regex.