Major security issue?

[Keith]

I have found what I believe to be a serious security
issue in ASP.Net. If you have:

1. Your website configured for anonymous access
2. Elect under web.config to set the sessionstate
attribute of cookieless to true

Anyone from any IP address or across another browser can
copy the URL and work within the session. My question
is "Why doesn't ASP.Net provide an option around ensuring
all requests for a user session originate from the same
IP address and/or same useragent?" I know that some
people sit behind firewalls, proxies and layer 4 devices
that could load balance and affect HTTP traffic, but it
honestly escapes me why I can access my web application
on any machine inside or outside of my network with just
the sessionid in the URL from even different browsers.
There must be a way to control this in the
configuration. Am I alone in find this troubling?

 

[Ray at]

It seems to me that this would be listed as a predictable downside to using
cookieless sessions. Verifying IPs and/or user agents wouldn't be any real
way to avoid this, so it makes sense to me that this wouldn't be the default
behavior for asp.net to check that. And if it were to check it, where would
it store this info? In session variables? Hmmph.

 

[Keith]

This is predictable in an insecure product. I'm not
trying to act as if I discovered something new or Earth
shattering, but I am quite surprised there is nothing in
place in ASP.Net to protect user sessions from being
hijacked. It seems to me that the session IDs have been
problematic since ASP first came about. In ASP.Net they
are still for some reason handed out in a fashion that
means the same ID could be sent out to the same browser
even after a Session.Abandon(). It doesn't make sense
that Microsoft couldn't do something as simple as encrypt
the user agent and source IP into the session GUID if the
user wanted to lock the source and device of a request
down to a particular computer/network. From an
architectural standpoint I realize that this in itself
would add some overhead to IIS because every HTTP request
would have to be checked against a lookup, but with HTTP
keep-alives this check would only need to occur once on
the same connection. I also realize that someone could
use this to DoS a server by sending lots of HTTP requests
with random IDs that would have to decoded and matched up
against connections, but I'm sure that intrusion
detection systems could be made to deal with this issue.
The other option of course is to not use cookieless
sessions under the anonymous user configuration and rely
on an in-memory cookie which is obviously a little less
accessible. In either situation though, this seems like
an incredible option to not provide ASP users.

 

[Ray at]

This is predictable in an insecure product.

I can loosen all your lugnuts with a standard crossbar wrench, too. Does
that make your car an insecure product? If you think it does, use >=3 wheel
locks on each of your wheels.

I'm not
trying to act as if I discovered something new or Earth
shattering, but I am quite surprised there is nothing in
place in ASP.Net to protect user sessions from being
hijacked.

There is, real sessions, although that is arguable as well. And this has
nothing to do with what server-side technology you choose to use.

It seems to me that the session IDs have been
problematic since ASP first came about. In ASP.Net they
are still for some reason handed out in a fashion that
means the same ID could be sent out to the same browser
even after a Session.Abandon().

Even if this happened, would it matter? It'd still be a new session.

It doesn't make sense
that Microsoft couldn't do something as simple as encrypt
the user agent and source IP into the session GUID if the
user wanted to lock the source and device of a request
down to a particular computer/network.

That data is meaningless though. When you have 1000 computers created from
the same image all sitting behind the same firewall, for example.

From an
architectural standpoint I realize that this in itself
would add some overhead to IIS because every HTTP request
would have to be checked against a lookup, but with HTTP
keep-alives this check would only need to occur once on
the same connection. I also realize that someone could
use this to DoS a server by sending lots of HTTP requests
with random IDs that would have to decoded and matched up
against connections, but I'm sure that intrusion
detection systems could be made to deal with this issue.
The other option of course is to not use cookieless
sessions under the anonymous user configuration and rely
on an in-memory cookie which is obviously a little less
accessible.

Cookieless sessions are just an alternative. If you're that worried about
them, don't use them. This is not a design flaw in ASP; this is just a
result of the technology that you're using and the way it works. If you
choose to use querystrings to identify users, it doesn't matter what kind of
server-side technology you use if you're catering to the cookie-paranoid
people.

In either situation though, this seems like
an incredible option to not provide ASP users.

I keep having flashbacks to "You are already logged into another
workstation" messages from Novell clients after your computer blue screens.
I don't know why, but I am. It's really foolish to build something into a
product that can often give false positives. What you're suggesting would
have that potential.

 

[Paul Glavich]

I replied to this on the other list, but thought I'd send it here as well.

We have used cookieless sessions and what you say is true, but we used SSL
to encrypt traffic, which as you know requires a connection to the same
client/server (ie. if connection broken, then the SSL session is invalid) so
this IP verification approach could still work but it assumes SSL, which of
course is really outside of ASP.NET's domain.

Further to this you could use client certs to verify integrity which
strictly doesn't stop people from hjacking a session (simply minimises it),
but there are just som many ways to approach this, each with positives and
negatives, that if the ASP.NET team adopted one approach, it would be
implicitly be advocating this one approach which may very well be flawed
under a number of different situations.

My 2 cents.

 

[Max]

Yes, for testing purposes I've created a 10 line VB.NET program that simply
loops thru until it enters another user's session. It does work on a lot of
ASP.NET web sites that use cookieless sessions, but it's rather random and
you don't necessarily hit any confidential data. I'm not really sure what
the implications are. In any case, I don't use cookieless sessions for
authentication. lol!

-Max