All of my wrapper components have a conditional i.e. "$rootPage.isAuthorized" 
around the WOComponentContent.  Solves this issue and is handy in dozens of other ways.

On 4/10/12 7:14 AM, Patrick Robinson wrote:
I'm pretty sure this "feature" is the only mechanism by which a user can 
request a specific page (or component) by name.  I would want to block arbitrary access 
to pages as well as prevent spurious session creation.

But yes, there are ways to mitigate the effects.  If an authenticated "user" is 
stored in the Session, then you can check for that before performing an action in 
invokeAction() or returning a response in appendToResponse().  And you *do* have to worry 
about invokeAction(), by the way: the presence of a senderID in the URL causes the 
component action handler to initiate the invokeAction phase.  I suppose sessions with no 
authenticated user could even be terminated at the same time.

No end to the fun!

- Patrick

On Apr 10, 2012, at 2:43 AM, Cheong Hee (Gmail) wrote:

Hi Patrick

The rationale I am asking is the way web technology is, I think we may not be 
able to block the arbitrary access of web pages.  However, if we could use user 
authentication as a way to check, terminate the unwanted sessions and redirect 
to another stateless page, the impacts could be reduced. Correct me if wrong..

Cheers

Cheong Hee

----- Original Message ----- From: "Cheong Hee (Gmail)"<[email protected]>
To: "Patrick Robinson"<[email protected]>
Cc: "WebObjects-Dev Mailing List"<[email protected]>
Sent: Tuesday, April 10, 2012 12:53 PM
Subject: Re: preventing direct component access


Hi Patrick

This is an interesting old issue.  Just curious, what will be your ultimate 
ideal resolution to this?  Bar the access of the page, or reduce the redundant 
sessions creation or something else ...

Cheers

Cheong Hee

----- Original Message ----- From: "Patrick Robinson"<[email protected]>
To: "Amedeo Mantica"<[email protected]>
Cc: "WebObjects-Dev Mailing List"<[email protected]>
Sent: Tuesday, April 10, 2012 4:52 AM
Subject: Re: preventing direct component access


That code represents the per-app version of the "conventional wisdom" that I started out 
questioning, below.  The problem with this is that the user can specifiy a "senderID" (as in the 
URL I gave there), and then senderID() will *not* return null; in the case below, it'll be "99".


On Apr 9, 2012, at 4:48 PM, Amedeo Mantica wrote:

Try this in your Application.java:

public WOComponent pageWithName(String pageName, WOContext context)
{
if((context.senderID()==null)&&(componentRequestHandlerKey().equals(context.request().requestHandlerKey())))
{
log.error("Direct Access attempt");
pageName="Main";
}
return super.pageWithName(pageName, context);

}



On 09/apr/2012, at 21:59, Mike Schrag wrote:

Yeah, you're right ... might be kind of a pain in the butt to fix without 
hackery then :)

On Apr 9, 2012, at 3:41 PM, Patrick Robinson wrote:

But it doesn't even have to have the ".wo" on the end of the page name for this hack to 
work.  If the app has a "SecretPage.wo" component, then a URL like this will instantiate 
and return it:

https://myhost.mydomain/cgi-bin/WebObjects/MyApp.woa/wo/SecretPage//88.99

- Patrick


On Apr 9, 2012, at 10:10 AM, Mike Schrag wrote:

probably just catch any time you have a ".wo" in your URL and throw ... you 
could do it in the url rewriter or something. i don't think there's ever any reason to 
have a .wo reference in a normal app.

ms

On Apr 9, 2012, at 10:00 AM, Patrick Robinson wrote:

Yeah, that _does_ sound rather annoying!  :-P

Is there a perhaps less-annoying way to approximate similar behavior?


On Apr 5, 2012, at 2:46 PM, Mike Schrag wrote:

I changed this in WO core, and unfortunately it's kind of annoying to fix without some 
hackery, but in WOComponentRequestHandler, there's a static method 
requestHandlerValuesForRequest ... That dictionary has a key named "wopage" in 
it. If you did some class rewriting (with like gluonj or something), you could change 
that static method to remove the wopage key ... That MIGHT be enough to do it.

On Apr 5, 2012, at 2:39 PM, Patrick Robinson wrote:

I've stumbled across a wrinkle re: what I had assumed to be the conventional 
wisdom for preventing direct access to component pages via URLs like the 
following:

http://myhost.mydomain/cgi-bin/WebObjects/MyApp.woa/-9876/wo/SecretPage.wo

It's an old, old WO problem, and I'm wondering what other people do to handle 
it.

I've always figured the best idea is to just configure the web server to catch 
WO URLs that end in /wo/(.+)\.wo and rewrite or redirect them.  Another 
potential approach is to try to recognize and catch such requests in the app 
itself, somewhere like the Application class's pageWithName.  The problem is, 
these solutions don't catch all the sneaky ways of slipping in a back door.

Consider:

http://myhost.mydomain/cgi-bin/WebObjects/MyApp.woa/-9876/wo/SecretPage.wo//1.2

This ends up with Application's pageWithName trying to create a page with the name 
"SecretPage".  A new session has already been created somewhere down inside the 
component request handler, it'll have a WOContext with a contextID of 0, and the senderID 
will be 2. You'd be hard-pressed to know that you shouldn't allow the page creation to 
proceed.

You could try to change the web server's search pattern to also catch a slash followed by more 
characters after the ".wo", but you'd have to be careful not to disallow sessionIDs that 
just happen to end in "wo".  And even if you could reliably block the above, the hacker 
could try this:

http://myhost.mydomain/cgi-bin/WebObjects/MyApp.woa/-9876/wo/SecretPage.wox//1.2 (that 
is, add more characters after the ".wo")

Now that doesn't fit the pattern at all, and gets hung up in the Application's 
pageWithName, where a way-too-informative WOPageNotFoundException is thrown.  
Of course, you'd catch that somewhere like handleException().  Doesn't quite 
seem like the right approach, either.

My point here is, there are more ways of hacking a WebObjects URL than I had 
previously considered.  Does anyone have what they consider to be an ironclad 
solution to this problem?

(I hate it when I discover stuff I thought I had dealt with 10 years ago is 
still biting me.)

- Patrick


_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list      ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/webobjects-dev/mschrag%40pobox.com

This email sent to [email protected]






_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list      ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/webobjects-dev/amedeomantica%40me.com

This email sent to [email protected]



_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list      ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/webobjects-dev/chng34%40gmail.com

This email sent to [email protected]




  _______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list      ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/webobjects-dev/lists%40lochgarman.com

This email sent to [email protected]


_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list      ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com

This email sent to [email protected]

Reply via email to