Security with UrlScan
UrlScan is an invaluable security tool for Web sites. Although IIS 6.0 incorporates some of its functionality, it’s still useful for IIS 6.0 sites—and critical for older versions of IIS.
You can take a lot of steps to help protect your Internet-facing machines
from malicious attacks. The most fundamental is to keep on top of your
Web server configuration and make sure you’re not accidentally opening
up a Web site area to potential abuse. Utilities like the IIS Lockdown
tool and many of the built-in security features of IIS 6.0 will certainly
help in the effort.
The other chief method is through implementation of UrlScan. UrlScan
does exactly what its name suggests: It scans incoming URL requests and
its associated data. It uses a series of rules to determine whether the
information in that request is potentially dangerous (i.e. requesting
an application or accessing a folder with periods in its name), or contains
information not normally expected (such as high characters—UTF8 or Unicode—or
excessive amounts of information). To help you diagnose any potential
problems and any attempts to upset your server, UrlScan can also log requests—including
the offending request data—up to a 128KB for each request.
Because UrlScan works as a filter before the information is passed on
to the script or application that handles the request, it can act as a
buffer, so you don’t have to modify your existing code. Therefore, if
a request is identified as being a potential risk, the script can immediately
return a 404 error (its own or the IIS custom error message) to the client,
without the information ever reaching the script. This protects the script,
your Web site and your server.
UrlScan isn’t new. Version 1.0 was released by Microsoft in 2001. Version
2.0 came out soon after to fix some early bugs and problems in the first
version that actually broke many Web sites. Version 2.5 was released by
Microsoft to coincide with the release of IIS 6.0, which is part of Windows
IIS 6.0 contains some significant security improvements of its own, including
providing some of the functionality of UrlScan, but that doesn’t mean
that IIS 6.0 replaces the tool. In general, I recommend installing UrlScan
on all IIS servers that provide some sort of dynamic content or host any
type of application. Even if your site only contains static data, installing
UrlScan can help to track potential attempts to break into your servers,
which helps prevent future attacks and provides useful information on
other areas of the network that were potentially infiltrated. And, of
course, there are still many Web sites out there running previous versions
of IIS. Those sites should consider UrlScan a mandatory tool. (Microsoft
recommends only using UrlScan 2.5 with IIS 6.0; users of previous IIS
versions should use 2.0.)
Download the UrlScan installer from the Microsoft Web site. You can get
it from UrlScan homepage at www.microsoft.com/technet/security/urlscan.asp.
The installer adds the UrlScan.dll and UrlScan.ini configuration file
to the %systemroot%system32\inetsrc\urlscan directory. If you’re upgrading
to UrlScan 2.5 from a previous version, your existing configuration information
won’t be overwritten, but it will have been updated to configure the default
settings for the new features.
UrlScan will be added to IIS as a global ISAPI filter, applied to all
of your Web sites. If you only want it to apply to some of your Web sites—perhaps
because it’s broken some of your sites until you can resolve the issues—remove
it from the global filters list and add it to each site individually.
If you do this, make sure to promote UrlScan to be the first ISAPI filter,
to ensure the maximum protection.
Within IIS 6.0, UrlScan also makes some configuration changes to ensure
that the logging and execution of the script work correctly with the new
process-based model of the IIS 6.0 worker process isolation mode.
Configuration and Operation
The UrlScan extension uses the parameters in the urlscan.ini file installed
in the %systemroot%\system32\inetsrv\urlscan folder to determine which
requests to let through, and which to reject. There are a number of different
options; as with any security configuration, it’s a question of balance
between protecting your servers and network and providing enough access
to allow the application and system to provide necessary services.
The configuration file takes the form of a standard .ini file and is
logically split into two sections. The primary section, Options, specifies
a combination of global options with references to specific elements within
the .ini file to process within the logical second part of the file. For
example, the UseAllowVerbs option specifies whether UrlScan should process
the [Allow Verbs] section or the [DenyVerbs] section. Figure 1 shows a
small sample of the configuration file.
|Figure 1. The UrlScan configuration file is a
basic .ini file, specifying all the different base options and main
sections within the rest of the file to be processed. (Click image
to view larger version.)
UrlScan intercepts requests (by acting as the primary ISAPI filter).
It uses the options within the configuration file to determine whether
the request is valid. If the request passes the qualification, the ISAPI
filter lets the file go on to the next ISAPI filter in the list. If rejected,
UrlScan redirects the request to an error page.
You can control the specific response behavior in the event of a rejection
through the configuration file. The UseFastPathReject option, if set,
forces the request to be rejected immediately, in which case the user
gets a generic 404 error message. The request itself isn’t logged in the
UrlScan log, which is useful if performance is critical. If that option
isn’t set, UrlScan redirects the user to the URL pointed to by the RejectResponseUrl
Rather than looking at the rest of the options in detail, we’re going
to concentrate on how UrlScan integrates with, or has been replaced by,
Augmenting IIS 6.0
There are two ways in which UrlScan augments the security within IIS 6.0.
The first is by monitoring the size of incoming requests (URL length,
HTTP header length and query string data) and rejecting them before the
information is ever sent to the script or application. The second augmentation
is in the filtering of the URL requested from the server. Configurable
parameters for size monitoring include:
The [RequestLimits] section of the configuration file specifies the limits
contained in a request. Any request exceeding these limits is automatically
MaxAllowedContentLength is the maximum size of the Content-Length header
field, which stores the length of any additional data (such as that from
a form) or files uploaded through WebDAV. The default is 30,000,000 bytes,
which—let’s face it—is quite high. Even working with fairly hefty WebDAV
requirements, it’s unlikely you’d be posting documents that size. If your
sites use relatively light forms and don’t support WebDAV, you might want
to set this as low as 1,000 bytes. Even with WebDAV, I’d recommend reducing
this figure to 100,000.
MaxUrl is the maximum size of a URL request. The default is 26.00 (the
same length as the maximum size of a path) and should stop clients from
trying to access excessively long URLs. Generally, you can leave this
figure at its standard level, but be aware that heavily organized and
directory-led sites might find this limitation restrictive.
MaxQueryString is the maximum length of a query string appended to a URL.
The default is 2,048 bytes, which should be more than enough to handle
a simple form. You can also limit specific HTTP headers by prefixing the
header with “Max-”. Using these requires in-depth knowledge of the HTTP
headers sent in a typical request. If you want to make use of this, I’d
set the following limits:
The URL filtering system rejects requests based on URL content, including
some you’d never have thought were allowed in the first place. The primary
configuration options here are:
AllowHighBitCharacters. If set to 1, this allows characters with an ASCII
or Unicode number above 128. On a strictly English-based site it’s probably
best to enable this. On a site with foreign language material, you might
find that this blocks perfectly legal pages.
AllowDotInPath. If set, this allows one or more periods in a URL (outside
of a file extension). Enabling this should stop users from trying to refer
to parent directories by using the “..” notation in a directory.
[DenyUrlSequences]. This specifies a list of characters to be rejected
in a URL. The default options here are “..”, “./”, “\”, “:”, “%” and “&”.
You might also want to add “,” , “$”, “^”, “!”, “@”, “~”, “<>”.
IIS 6.0 has certainly gotten better at working with URLs and components
it doesn’t recognize—now, it just ignores them—but an extra level of filtering
will help address issues you might not have considered yet. Of course,
the flip side to adding these filters is to ensure that your Web team
doesn’t develop sites whose URLs include those banned characters.
Using IIS 6.0 instead of UrlScan
The main focus for Microsoft when developing IIS 6.0 was to improve the
overall security of its Web server. This starts with IIS no longer being
installed by default. It continues through all areas of the system, including
the mechanisms for restricting files, extensions and applications, as
well as worker process isolation.
Does it therefore make UrlScan obsolete? No. But there are quite a few
areas where UrlScan previously provided security that IIS 6.0 is now capable
of handling. A few of the key areas IIS 6.0 addresses that supercede the
functionality offered by UrlScan include:
Extensions. In past versions of IIS you could access any file, providing
you knew its full name and extension. UrlScan and the DenyExtensions feature
could be used to allow access only to files with recognized extensions.
In IIS 6.0, files are only returned if IIS can identify the file’s MIME
type from the MIME type/extension mapping built into the server. If an
extension isn’t in the map, there’s no access to the file. I always recommend
switching off all MIME types within IIS, except those required for the
sites you’re serving. For example, in a basic, static site you might limit
the MIME types to support only .htm, .gif and .jpg files. You can see
a sample of the MIME type properties in Figure 2.
|Figure 2. You can set MIME types on a server-
or Web site-level. Unlike previous IIS versions, IIS 6.0 will only
honor requests for files it can identify through the MIME type map—any
other file/extension not listed generates a 404-class error. (Click
image to view larger version.)
WebDAV. In the past, UrlScan could help control WebDAV access; now you
can control WebDAV access directly through IIS. To enable or disable WebDAV
for your server, use the Web Service Extensions Manager, available within
the IIS tree of IIS Manager.
URL filtering. IIS 6.0 provides basic URL filtering by changing the way
it determines whether or not to execute an application or script. This
makes the NormalizeUrlBeforeScan option of UrlScan irrelevant, as the
URL is no longer the sole determiner of whether to execute a script. Also,
the AllowDotInPath option has been superceded by the same execution mechanism
and the improvements to the MIME map that limits which files are returned.
Request lengths. Although not heavily promoted, you can limit some elements
of a request directly within IIS 6.0. There are two areas where you can
set limits: Through the MaxRequestEntityAllowed and ASP MaxRequestEntityAllowed
Metabase properties, and a set of Registry properties within the HKEY_LOCAL_
MACHINE\System\CurrentControlSet\ Services\HTTP\Parameters node.
Even with these enhancements, UrlScan has a vital role to play in the
security of your server. If you don’t want to use UrlScan, at least make
sure you’ve made the changes suggested above. For extra protection, use
UrlScan to help IIS 6.0 make more informed decisions about what URLs and
pages it should be serving.
Although UrlScan and IIS 6.0 combined are an effective way of protecting
yourself from a large number of attacks, they shouldn’t be your only means
of protection. You should also install and consider at least two or more
of the following:
Firewalls still play a vital role in the protection of your site and servers.
The primary purpose and advantage of a firewall is to protect and limit
access to your network to and from specific IPs and through specific ports.
A firewall is, in fact, the only effective way to restrict traffic through
your site to the types you need to support. A good firewall will also
monitor for denial of service (DoS) attacks and port scans and will include
the ability to block off connectivity from hosts or sites you suspect
may be launching attacks.
IPSec/VPN technology is best applied within your company network, to protect
the security of the machines using your intranet, and for companies and
organizations using applications remotely. For example, an extranet or
remote office should use IPSec and/or a traditional VPN solution to ensure
that communication and access between that office and your machines is
encrypted and secured.
Bad code is always a worry. There is, of course, no way to guarantee or
protect yourself from bad code, but UrlScan certainly goes a long way
toward addressing some of the problems that bad code can highlight. You
should always test your code, including stressing it by over- and under-supplying
information and additional data not normally required or accepted by the
Keeping up to date should really be top of the list. Don’t underestimate
the significance of that security announcement from Microsoft or the latest
security patch. You should always install these patches on your test machines
once they’ve been released and ensure it doesn’t break any of your existing
Finally, don’t ignore common sense. All the technical security in the
world, including UrlScan, won’t save you if you forget to use decent passwords,
close off ports and features not really needed, or leave the door to your
network operations center propped open with a fire extinguisher.