Xymon 4.3.x Buffer Overflow / Code Execution / Information Disclosure
Posted on 16 February 2016
Hi, Multiple security issues have been found in the server component of the Xymon monitoring system. These issues affect all versions of Xymon 4.3.x prior to 4.3.25, as well as the obsolete 4.1.x and 4.2.x versions. All issues have been resolved in Xymon 4.3.25, released on Feb 8 2016. It is available at http://sourceforge.net/projects/xymon/files/Xymon/4.3.25/xymon-4.3.25.tar.gz/download Details ------- CVE-2016-2054: Buffer overflow in xymond handling of "config" command. The xymond daemon performs an unchecked copying of a user-supplied filename to a fixed-size buffer when handling a "config" command. This may be used to trigger a buffer overflow in xymond, possibly resulting in remote code execution and/or denial of service of the Xymon monitoring system. In the case of remote code execution, it will run with the privileges of the xymon userid. This bug may be triggered by anyone with network access to the xymond service on port 1984, unless access has been restricted with the "--status-senders" option (a non-default configuration). CVE-2016-2055: Access to possibly confidential files in the Xymon configuration directory The xymond daemon will allow anyone with network access to the xymond network port (1984) to download configuration files in the Xymon "etc" directory. In a default installation, the Apache htaccess file "xymonpasswd" controlling access to the administrator webpages is installed in this directory and is therefore available for download. The passwords in the file are hashed, but may then be brute-forced off-line. This bug may be triggered by anyone with network access to the xymond service on port 1984, unless access has been restricted with the "--status-senders" option (a non-default configuration). Administrators should verify this file is not readable by the xymon user and modify ownership and permissions as needed. Additionally, the following restrictions have been added to files requested via "config" messages sent to xymond: * They must be regular files as returned by stat (no symlinks) * They must end in ".cfg" The restriction on file names ending in ".cfg" can be overridden by setting ALLOWALLCONFIGFILES="TRUE" in xymonserver.cfg and restarting xymond. Note that config files are processed through normal xymon file reading, so features such as "include" and "directory" still work when retrieving files over the network. These included files are not subject to the same restrictions. CVE-2016-2056: Shell command injection in the "useradm" and "chpasswd" web applications The useradm and chpasswd web applications may be used to administer passwords for user authentication in Xymon, acting as a web frontend to the Apache "htpasswd" application. The htpasswd command is invoked via a shell command, and it is therefore possible to inject arbitrary commands and have them executed with the privileges of the webserver (CGI) user. This bug can only be triggered by web users with access to the Xymon webpages, who are already authenticated as Xymon users. However, when combined with CVE-2016-2055 which allows for off-line cracking of password hashes, this bug may be exploitable by others. CVE-2016-2057: Incorrect permissions on IPC queues used by the xymond daemon can bypass IP access filtering An IPC message queue used by the xymon daemon is created with world-write permissions, allowing a local user on the Xymon master server to inject all types of messages into Xymon, bypassing any IP-based access controls. Exploitation of this bug requires local access to the Xymon master server. CVE-2016-2058: Javascript injection in "detailed status webpage" of monitoring items A status-message sent from a Xymon client may contain any data, including HTML, which will be included on the "detailed status" page available via the Xymon status webinterface. A malicious user may send a status message containing custom Javascript code, which will then be rendered in the browser of the user viewing the status page. Exploitation of this bug requires that you can control the contents of a status message sent to Xymon, which is possible if you control one of the servers monitored by Xymon, or the Xymon master server. Also, the bug requires a user to actually view the "detailed status" webpage. This bug has been patched in Xymon 4.3.25 by including a "Content-Security-Policy" HTTP header in the response sent to the browser. This means that older browsers may still be vulnerable to this issue. CVE-2016-2058: XSS vulnerability via malformed acknowledgment messages (Note that this uses the same CVE id as the Javascript injection issue) The message sent by a user to indicate acknowledgment of an alert is not HTML-escaped before being displayed on the status webpage, which may be used to trigger a cross-site scripting vulnerability. Exploitation of this bug requires that the attacker is able to acknowledge an alert status. This requires user-authenticated access to the Xymon webpages, or that the user receives a message (usually via e-mail) containing the authentication token for the acknowledgment. Credit ------ We would like to thank Markus Krell for reporting these issues, and for working with us to resolve them. On behalf of the Xymon team, Henrik Størner
