fun with the web and security

[daw@orodruin.CS.Berkeley.EDU (David A Wagner)]

Here's a fun way to exploit security holes via the web:
	http://www.cs.berkeley.edu/~daw/js1.html
A rough representation of its contents follow.



Whee! The web is awfully convenient for exploiting security bugs.... 

The following URL contacts your sendmail SMTP server and attempts to exploit
an old, well-known security hole, trying to gain root access. Click _here_
to try it. 

As it stands, clicking on the URL above does not do anything harmful to your
machine-- but it could! (This is a test of the emergency broadcast system.
This is only a test.) 
______________

We can get you to send arbitrary text, to an arbitrary port on an arbitrary
host, from your machine.  (If you are inside a firewall, we can thereby send
arbitrary text to any internal machine by getting you to click on the link
above.) The technique is simple: we list the host and port in a gopher URL,
and encode the text to be sent in the path. 

For instance, a successful exploit of the hole could leave a backdoor root
shell, and inform us via a pseudonym at an anonymous remailer. 

The exploit could be hidden by use of the JavaScript "width=1,height=1"
techniques pioneered at John LoVerso's _JavaScript security hole page_; then
you wouldn't even know when you'd been attacked. 

The exploit could be forced on you via many standard tricks: the Redirect:
or META-EQUIV Refresh: or JavaScript mechanisms work fine, for instance. 

This is most dangerous when you are behind a firewall. Typically, there will
be many machines inside a firewall which run insecure software. Normally,
that would be safe, since the firewall prevents an outsider from connecting
to the unsafe sendmail servers inside-- yet the example URL above allows
outsiders like us to exploit security holes on the inside of your firewall.
Nothing stops us from putting the IP address of a vulnerable machine inside
your firewall in the URL above, and waiting for you to click on it: the
firewall doesn't prevent connections from you to the internal vulnerable
machine, and thus can't stop this attack. Using JavaScript, we don't even
have to wait for you to click on anything. Furthermore, a JavaScript program
could systematically and invisibly try all the machines inside your firewall. 

We could have used many other well-known security holes: there's nothing
special about this particular sendmail bug (except that it was convenient
for us to implement). 
______________

Be afraid. Be very afraid. 
-- Ian Goldberg and David Wagner.