|
COMMAND Log corruption via specially crafted reverse DNS data SYSTEMS AFFECTED WebExpert, LoganPro, Iplanet Log Analyzer, WebTrends, SurfStats, WebLogExpert PROBLEM In Hugo Vázquez Caramés & Toni Cortés Martínez of Infohacking Research 2001-2003, whitepaper : ILLC - Inverse Lookup Log Corruption We are using a technique that we have called “ILLC” (Inverse Lookup Log Corruption) that allows us to corrupt the logs generated by many web servers that are doing inverse address resolution. Impact of this technique: - “IP spoofing” on the logs - Code execution (XSS) on boxes that are running log analyzers (web servers that have buit-in report analisys tools,etc.) On some specific scenarios, we have been able to hide the entire http request to the log viewer. Most of the actions were possible because of the lack of filtering when parsing host names between different applications. Related RFC´s about Internet Host Names convention: RFC 952: “(…) 1. A "name" (Net, Host, Gateway, or Domain name) is a text string up to 24 characters drawn from the alphabet (A-Z), digits (0-9), minus sign (- ), and period (.). Note that periods are only allowed when they serve to delimit components of "domain style names". (See RFC-921, "Domain Name System Implementation Schedule", for background). No blank or space characters are permitted as part of a name. No distinction is made between upper and lower case. The first character must be an alpha character. The last character must not be a minus sign or period.... Single character names or nicknames are not allowed.... (…)” RFC 1034: “(…) 3.5. Preferred name syntax ... The labels must follow the rules for ARPANET host names. They must start with a letter, end with a letter or digit, and have as interior characters only letters, digits, and hyphen. There are also some restrictions on the length. Labels must be 63 characters or less. (...)” RFC 1123: “(…) 2.1 Host Names and Numbers The syntax of a legal Internet host name was specified in RFC-952 [DNS:4]. One aspect of host name syntax is hereby changed: the restriction on the first character is relaxed to allow either a letter or a digit. Host software MUST support this more liberal the RFCs. Note that under BIND 8, you may need to add "check-names master ignore" to the zone definition when defining these names.(…)” RFC 2181: “(…) 11. Name syntax Occasionally it is assumed that the Domain Name System serves only the purpose of mapping Internet host names to data, and mapping Internet addresses to host names. This is not correct, the DNS is a general (if somewhat limited) hierarchical database, and can store almost any kind of data, for almost any purpose. The DNS itself places only one restriction on the particular labels that can be used to identify resource records. That one restriction relates to the length of the label and the full name. The length of any one label is limited to between 1 and 63 octets. A full domain name is limited to 255 octets (including the separators).(…)” Independently of what should be the legal host name syntax, it seems that operating systems allows host names with arbitrary characters. To succesfully attack a server with “ILLC” technique is mandatory that web server/log analyzer,etc., will be doing inverse address resolution and that the attacker could control in any way the responses to those inverse lookup requests. --------------------------------------------------------- Exploiting web server/log analyzers through “ILLC” --------------------------------------------------------- Examples of attacks: -Log “IP Spoofing” (exploited succesfully on Apache 2.0.44 on Windows/Linux, and Iplanet 6 on Windows) Scenario: a machine with a host name as "123.123.123.123" makes a request to an Apache server. If the server dosn`t generate any error, on the access log you will see an access request from a client called "123.123.123.123", what apparently seems to be a valid request from a client that server was unable to resolve to a host name. So the real IP wouldn't appear in the access log file. access.log 123.123.123.123 - - [28/Feb/2003:10:39:01 +0100] "GET / HTTP/1.1" 200 1786 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0)" 123.123.123.123 - - [28/Feb/2003:10:39:46 +0100] "GET /badrequest.html HTTP/1.1" 404 294 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0)" If the request produces some error, you will see an entry in the error log file were you could see the real IP, although the web server has the inverse lookup activated. error.log [Fri Feb 28 10:39:46 2003] [error] [client 172.26.50.45] File does not exist: C:/Archivos de programa/Apache Group/Apache2/htdocs/badrequest.html So, while there aren’t errors, the real IP is not showed. This can lead in a complete anonymous http access for a client in a usual web surfing activity, that is, if there are not broken links,etc. In the case of Iplanet 6, the real IP wouldn’t appear in the “access” preview (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-access-log1.gif Neither in the “errors” preview (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-errors.gif -CODE INYECTION (Succesfully exploited on Apache 2.0.44 on Windows/Linux, on IIS 6.0 and Iplanet 6 on Windows) Scenario: a machine with a hostname as “<scrip>alert(‘a’)</script>” that makes an HTTP request leaves javascript code on the log. When generating a report, with some log analyzers (that show results in html), the script will be executed. *Note: in IIS 6.0 case we needed to restrict access on webserver by domain name in order to force inverse lookup resolution. *Note2: in the Iplanet case we needed to simulate a FQDN client host name like this: “<scrip>alert(‘a’)</script>.infohacking.com”. You can also set a host name were the script is only part of the entire string label: “nop<scrip>alert(‘a’)</script>.infohacking.com” so when html formatted it will appear as a valid domain name: “nop.infohacking.com.” Meanwhile the script will be executed… Some log analyzers proved to be vulnerable to "ILLC": WebTrends (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-webtrends-illc.gif SurfStats (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-surfstats_loganalizer.gif WebLogExpert (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/weblogexpert_illc.gif And probably many more… Iplanet comes with a buil-in tool to generate html reports based on access and error logs. This tool is part of the administration web interface. Moreover, Iplanet log analyzer always uses a web broser to show the results of the report, although the user selects “Only text output”, so it will be always exploitable. Iplanet Log Analyzer (HTML report, see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-report-html.gif Iplanet Log Analyzer (“txt” report, see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-report-text.gif On the other hand we have to notice that the access log previewer(“View Access Log”) in the Iplanet web interface is doing some kind of filtering on some characters (for example <>). Iplanet “View Access Log” (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-iplanet-filtra.gif -HIDING REQUESTS (Iplanet 6 on Windows) In the specific case of Iplanet 6, we coul realise that there`s a way to trick the server on not showing the request in the log preview (“View Acces Log” and “View Error Log”). The requests from boxes whose host name begins with “format=” will not be showed, that is, those requests still are visible in the access and error log files, but they would be “invisible” from the built-in access and error log viewer of the administration web interface…(“Last 25 accesses to…”). As an example we made requests from a box with this host name: “format=.infohacking.com”, and we realize that we can see the request in the access log: format=%Ses->client.ip% - %Req->vars.auth-user% [%SYSDATE%] "%Req- >reqpb.clf-request%" %Req->srvhdrs.clf-status% %Req->srvhdrs.content- length% "%Req->headers.referer%" "%Req->headers.user-agent%" %Req- >reqpb.method% %Req->reqpb.uri% %Req->reqpb.query% "%Req->reqpb.protocol%" %vsid% format=.winmat.com - - [28/Feb/2003:10:22:25 +0100] "GET /evilrequest.html HTTP/1.1" 404 292 "-" "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0)" GET /evilrequest.html - "HTTP/1.1" https-script.winmat.com But on the “View Access Log” nothing is showed (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-NOlog-iplanet.gif We suppose that server is processing the first part of the host name string “format=” as it would be the directive that sets the log format, and the rest of the string is not recognized as valid format, so nothing is showed. Combining the possibility of hiding a request and the Cross Site Scripting technique we could execute scripts on the machine that runs the Report Generator of the Iplanet Web Server in a “stealth” way. We have done this establishing a host name like this: “format=<script>alert(document.cookie)</script>.infohacking.com” (See link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-iplanet-cookie.gif Many more evil actions can be done... it only depends on the attacker's imagination. We haven’t checked “ILLC” on other daemons as ftp, smtp, or firewalls, IDSs, etc. We think that probably this technique could be used in the same way in other scenarios. --------------------------------------------------------- Exploiting http headers for log corruption --------------------------------------------------------- Controlling inverse lookup responses is not always possible for the attacker. We tried to figure out another, more generic attack to corrupt web logs. The first that came to us was to use faked http headers in order to achieve the same result: execution of scripts by log analyzers. There are a lot of http headers that can be used to inject code in a log file. We are not going to discuss all of them in this paper, but only to outline some generic ways to do it. The main objective here is to choose the right header to inject code in the http request… For example, the “RequestResource” is always showed in web logs, but probably it will be filtered by many application firewalls or it will be detected by IDSs… On the other hand the “UserAgent” header usually is not being checked for suspicious secuence of characters, and web masters usually like to have this info on their log files… An example on how to trick a log analizer to execute a script we set “UserAgent” header of our client as: “<script>alert(‘UserAgent’)</script>”. The requests of our client with this faked “UserAgent” will inject code in the web server log. Some log analyzers reading this logs and generating HTML formatted reports without filtering the output, will execute the script. -Examples of vulnerables log analyzers- WebExpert (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/log_anal_XSS.gif LoganPro (see link below): http://www.infohacking.com/INFOHACKING_RESEARCH/Our_Advisories/ILLC/cap-loganpro-agent.gif To solve this kind of problems it would be nice a more aggressive filtering on DNS responses and HTTP requests on all the headers. To finish this short analisys we would like to make some questions: Are log analyzers thrusting too mutch on log files? Maybe, are web servers the ones that would have to filter what they write to log files…? Is the operating system the one that have to filter the returned values from DNS servers? Are the actual legal domain name hosts allowed too mutch liberal? SOLUTION See above