NAME Apache::Access::Headers - mod_perl HTTP header authorization module SYNOPSIS # in httpd.conf PerlSetVar HeadersAccessConf conf/headers_access.conf DocumentRoot /usr/local/apache/htdocs <Directory "/usr/local/apache/htdocs"> PerlModule Apache::Access::Headers PerlAccessHandler Apache::Access::Headers </Directory> DESCRIPTION This module is intended to be used as a mod_perl PerlAccessHandler. It's function is to authorize requests for server resources based on the existence of and content of HTTP headers. Authorizing HTTP headers may be be set by a web browser, a software agent, or an authenitcating proxy server. This module was originally written to work with the latter. Note: The default reponse from the handler is currently FORBIDDEN. This behavior is not yet configurable. CONFIGURING APACHE Module configuration is simple ( read: limited ). Currently, the module only works with a single configuration file, and works best when configured for a server's document root. See the LIMITATIONS section for an explanation of the modules current short-comings. Add the following line to httpd.conf outside all Directory, Location and VirtualHost blocks: PerlSetVar HeadersAccessConf /path/to/conf/headers_access.conf And add the following lines to the DocumentRoot Directory block: PerlModule Apache::Access::Headers PerlAccessHandler Apache::Access::Headers CONFIGURATION FILE General Options Although the modules is currently limited to a single xml-based configuration file, this configuration file is quite flexible. The shell of the conf file is: <headers_authz> <headers> [...] </headers> </headers_authz> The important part of the conf file is the <header> blocks within the <headers> block. Each <header> block must contain two items: an <id> tag and a <path> tag. The <id> tag specifies the name of the HTTP header that that must be set to allow access to the urls matched by the <path> tags. <path> tags are treated as regular expressions ( i.e., m|^$k$| where $k is the value of the <path> tag ). Using the Sample Configuration File below, a request for /secrets/index.html must contain an X-Can-View-Secret-Stuff header with a non-zero value in order to be successfully authorized. Likewise, a request for /secrets.html requires that either an X-Can-View-Secret-Stuff header or an X-Can-View-Super-Secret-Stuff header is present and set to a non-zero value. As mentioned above, <path> tags are treated as regular expressions. You'll notice, then, that the <path> tag for <id>X-Secret-User-I1 in the sample conf contains parantheses. Parentheses tells the module to require that the value assigned to the needed header ( i.e. X-Secret-User-ID ) equal $1. For example, using the sample conf, a request for /users/jeffo/ must have an X-Secret-User-ID header set to 'jeffo'. If X-Secret-User-ID header is present but set to 'tori', the request will be denied. Other Options There are three special configuration options. They are outlined here: ALL If the <id> of a <header> block is 'ALL', then _all_ requests for resources matched by the attached <path> tags. This is useful for allowing access to <path>/</path> and <path>index.html</path>, etc. REFERER If the <id> of a <header> block is 'REFERER', then the referer header is checked against an array of referer values specified by <referer> tags in the <header> block. For example, using the sample conf file, requests for /images/background.gif coming from a page on http://www.rulez.com/, or http://www.picnicman.com/ will be accepted. <referer> tags are treated as regexes ( like <path> tags ). The regex used is 'm|^$ar|' where $ar is the contents of the <referer> tags. One little trick is to use 'https?://' at the start of the regex to allow connections from either secure or insecure pages. Note: If the referer is not matched, FORBIDDEN will not be automatically returned. In this case, the module continues to loop over <path> values, looking for a secondary match. i.e. if a request for /images/button.gif does not come from a valid referer, but contains a X-Can-View-Secret-Stuff header, the image will be served. This behavior is open to debate. Admittedly, it's a hack meant to overcome some problems with the original spec. If you don't like it, then don't use REFERER as an <id>. :) <prefix></prefix> You'll notice a commented-out <prefix> tag in the sample conf file. The prefix tag was added because some - *cough* - authenticating proxies prepend a string to header values defined in the authtencation database. If the <prefix> tag is set, then all header checks will look for $PREFIX . $HEADER. For example, if <prefix> is set to 'Rulez-', requests for /secret/index.html would require not a X-Can-View-Secret-Stuff header, but a Rulez-X-Can-View-Secret-Stuff header. Sample Configuration File <header_access> <headers> <!-- <prefix>Rulez-</prefix> --> <header> <id>ALL</id> <path>/</path> <path>/index.html</path> </header> <header> <id>REFERER</id> <referer>https?://www.rulez.com/</referer> <referer>https?://ww.picnicman.com/</referer> <path>/images/.*</path> <path>/cgi/*.cgi</path> </header> <header> <id>X-Can-View-Secret-Stuff</id> <path>/secret/.*</path> <path>/secrets.html</path> <path>/images/.*</path> </header> <header> <id>X-Can-View-Super-Secret-Stuff</id> <path>/super-secret/.*</path> <path>/secrets.html</path> </header> <header> <id>X-Secret-User-ID</id> <path>/users/(.*?)/.*(</path> </header> </headers> </header_access> Configuration File DTD <?xml version="1.0" ?> <!DOCTYPE header_access [ <!ELEMENT header_access ( headers+ ) > <!ELEMENT headers ( header+ ) > <!ELEMENT header ( id, path+, referer+ ) > <!ELEMENT id ( #PCDATA ) > <!ELEMENT path ( #PCDATA ) > <!ELEMENT referer ( #PCDATA ) > ]> TESTING & DEBUGGING If you set Apache's LogLevel to 'debug', the module will spit out a bunch of information regarding it's handling of the request. ErrorLog ./logs/error_log LogLevel debug This is particularly useful if you're creating complicated <path> tag regexes, or if you really have no idea why a request is getting though. :) LIMITATIONS & THE FUTURE This module was originally written to cover the entire document root of a web server. And right now, that's all it does. This was a security-driven decision. The down-side is that the module is not VirtualHost or Directory friendly. I wanted to release this first version before adding support for use in VirtualHosts and Directory blocks with separate conf files. If interest - and time - allow, these will be the key features of the next release. ( Not to mention configuration of the default return policy and inversion of deny, allow rules, etc. ) RESOURCES For an example of a commercial authentication proxy, see SecureComputing's PremierAccess product. http://securecomputing.com/. I don't necessarily recommend it; I just know it because I wrote this to work with it. :) AUTHOR Jeffrey O'Connell, Jr. <jeffo@rulez.com> COPYRIGHT Copyright (c) 2003 Jeffrey O'Connell, Jr., Rulez New Media This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.