 |
WebGuard |
The web services have become a critical part of our computing infrastructure,
even our everyday life. Many application servers, script language, toolkits and
modules have been proposed to facilitate the quick construction of web services,
for example AOLServer, Apache, IIS, Tcl, PhP, ASP. The number of web sites
around the world increases exponentially, doubling roughly every six months.
Yet, this modular component approach fails to address security, a critical
cross-cutting concern in web services design. Usually, the web programmers need
to consider what security checks are appropriate and implement them
manually in site-specific code. However, this is highly error-prone. The
security policies and the implementations change frequently. Keeping the two up
to date and matched with each other is operationally difficult, and omission of
a check might result in a security breach. Our goal in WebGuard is to provide an
approach to formally specify security policies and enforce them automatically on
web services implementations. An automated mechanism for security
enforcement on web services include the following properties:
- Secure: The reference monitor implemented by the automatic
enforcement engine must correctly implement a give policy.
- Automatic: Security enforcement should not depend on manual
implementation by web programmers.
- General purpose: The language to express security policies should
be sufficient enough to express the common security concerns for web
services.
- Portable: The security enforcement mechanism should support various
different web service platforms.
- Backwards compatible: Existing web service implementation
should not need to be modified to incorporate the new automatic tool.
- Performance: The latency overhead of automatic security enforcement
should be comparable to manual code.
WebGuard is a mechanism for security enforcement on web services that
exhibits these properties. There are three contribution by WebGuard: (1) it separates
the high-level security policy from the implementation. (2) it provides a
general and versatile access control model for web services according to our
domain-specific language. (3) it demonstrates the overhead for automatically
generated security enforcement can be low.
WebGuard Overview